Root depth: a trait to increase water use and yield of wheat Alan David Severini A thesis submitted for the degree of Doctor of Philosophy of The Australian National University November 2015
Root depth: a trait to increase water
use and yield of wheat
Alan David Severini
A thesis submitted for the degree of
Doctor of Philosophy of
The Australian National University
November 2015
iii
Dedication
To Maru, Mario, Andrés, Marco and Joana,
who gave me the inspiration and strength needed
to overcome any obstacle.
v
Declaration
This thesis is an account of research undertaken between October 2011 and November
2015 in the division of Plant Industry, Commonwealth Scientific and Industrial
Research Organization (CSIRO), and the Research School of Biology, Australian
National University, under the supervision of Dr Michelle Watt, Dr Richard Richards,
Dr John Passioura and Professor John Evans.
The data presented in Figure 3.3 was adapted from a report presented by Dr Warren
Bond, Land & Water, CSIRO. Figure 6.1 was redrawn from Chenu et al. (2013 fig. 5),
and Figure 6.2 was adapted from Dardanelli et al. (1997).
Except where otherwise indicated, this thesis is my own original work.
Alan David Severini
vii
Acknowledgements
Thanks to my supervisors, Michelle Watt, John Evans, John Passioura, and Richard
Richards for their supervision and guidance during these four years of PhD. Thanks to
Anton Wasson, who wasn't within my PhD committee but behaved as a supervisor,
from whom I learned a considerable amount of science.
I'd also like to thank Gilbert, for excellent field assistance, practical ideas (always
smarter than mine), and most of all, his friendship. Mick provided courageous field
assistance facing any duty, like camping on site and measuring soil water with a neutron
probe once per hour during the whole night!
Everyone in our 'root team' made me feel that it was possible to do serious science while
having fun: Ritika, Vincent, Aki, Geetha, Harriet, Cathrine, Sarah, Kathy.
I am thankful to David Deery, José Jiménez-Berni, Wayne Thompson and Paul
Hutchinson, for help with all the work installing and getting data from gypsum blocks,
the infrared thermometers and other fancy devices, as well as sharing stimulating
discussions.
I am very grateful for all the amazing scientists at CSIRO Plant Industry, from whom I
learned a lot, sharing innumerable talks in the tea room, in the halls, or in the field.
My whole PhD field experiments would not been possible without the tremendous help
from the excellent team at Ginninderra Experiment Station: Phil Dunbar, Byron
Corcoran, Bernie Michelson, James Byrne, Tom McLachlan, and Matt Lynch.
viii
I am very happy of having met such friends as Leone Jansen and Peter Hairsine, for
their wise company, and giving me a free, wonderful Toyota Corolla 1989, that lasted
for the duration of the whole PhD and is still going!
Warren Bond gave me a lot of help with the neutron probe calibration data and
scientific teachings—like the Ockham's razor.
I thank also INTA, for letting me come to Australia to do the PhD and keeping my
position.
And surviving a PhD in Australia wouldn't be possible without the support from all my
close friends from Argentina, who always supported me in the distance.
I am greatly indebted with Australians and Australia, for letting me live and enjoy such
a beautiful country, which is indeed my second home.
And I deeply appreciate Katherine, for all her love and support when they were most
needed.
ix
Abstract
Crops with deeper roots could potentially capture more soil resources and as a
consequence yield more. However, as sampling roots by soil coring is challenging, there
are few examples of genetic diversity determined under field conditions. Canopy
temperature, an indicator of transpiration, could be used instead of direct coring to
screen for wheat varieties with increased access to deep water and hence deep roots in
the field. In this thesis we aimed (i) to seek genetic diversity in rooting depth, root
length density and relate these traits to yield in a wide range of triticale and wheat
germplasm, and (ii) to test the usefulness of continuously-monitored canopy
temperature and soil water status for phenotyping two commercial wheat varieties that
differ in rooting depth. In the first set of field experiments, rooting depth, root length
density and yield were measured in 34 wheat and 2 triticale varieties. Roots were
sampled by soil-coring with a tractor-mounted hydraulic press and were later counted
by the ‗core break‘ method. Root length density was predicted from root count density.
In the second set of experiments, canopy temperature was measured with fixed infra-red
thermometers, and soil water suction was determined with gypsum blocks buried at 20
cm intervals, from 20 to 160 cm depth. A crop water-stress index (CWSI) was
calculated to normalise for the effects of vapour pressure deficit over canopy
temperature. Soil water retention curves fitted to the soil of the site were used to convert
soil water suction into soil water content. Shoot biomass and grain yield were estimated
from 0.7 m2 samples per plot in all experiments. In the experiments seeking genetic
variability, we found that triticale produced deeper roots than commercial spring-wheat
(p < 0.10), and shorter varieties produced deeper roots than taller varieties (p < 0.10).
Moreover, rooting depth was related to shoot biomass (R2 = 0.66, p < 0.001) and grain
yield (R2 = 0.56, p < 0.001) across experiments and genotypes but not between
x
genotypes within the same experiment. In the experiments analysing canopy
temperature and water-use continuously, differences in deep-root length were not
statistically significant between the two varieties. The variety Gregory had greater root
length at depths beneath 1 m, was cooler, used more water and that water was
withdrawn from deeper soil layers than the other variety, Derrimut. Using CWSI gave
better predictions of soil water status than canopy temperature per se. By taking up
more water during grain filling, Gregory produced more yield at a rate of 54 kg ha-1
mm-1. CWSI did not correlate with day-to-day changes in water use. We conclude that
(i) there is genetic diversity in rooting depth within triticale and wheat germplasm; (ii)
by enabling the calculation of a CWSI, continuously measured canopy temperature
allows phenotyping of root systems with superior deep water access.
xi
Contents
Contents..................................................................................................................................... xi
List of Figures .......................................................................................................................... xv
List of Tables..........................................................................................................................xxiii
Chapter 1 Introduction: review of the potential of root depth to increase crop yield .........25
1.1 Introduction ...................................................................................................... 25
1.1.1 The framework of water- limited grain yield ............................................. 26
1.1.2 How can deeper root systems improve water- limited yield? .................... 33
1.1.3 Improvement of root depth through management and genotype selection
40
1.2 Thesis aims ....................................................................................................... 44
1.3 Thesis outline.................................................................................................... 44
Chapter 2 Field experiments used for wheat phenotyping..................................................47
2.1 Introduction ...................................................................................................... 47
2.2 Growing conditions: soils and weather ............................................................ 48
2.3 Sowing and agronomic management................................................................ 54
2.4 Germplasm sown .............................................................................................. 57
2.5 Statistical designs ............................................................................................. 64
2.6 Field measurements .......................................................................................... 71
2.6.1 Phenology .................................................................................................. 72
2.6.2 Soil coring to sample roots........................................................................ 72
2.6.3 Soil washing to separate roots from soil ................................................... 73
2.6.4 Root scanning to quantify root length density .......................................... 74
2.6.5 Predicting root length density from root count density............................. 76
2.6.6 Estimation of maximum rooting depth...................................................... 79
2.6.7 Root length per soil depth intervals or the whole soil profile ................... 81
2.6.8 Above-ground biomass and grain yield .................................................... 81
xii
Chapter 3 Using gypsum blocks for phenotyping water use in the field............................ 83
3.1 Introduction ...................................................................................................... 83
3.2 Measuring bulk density .................................................................................... 85
3.3 Estimating soil water content with neutron probe............................................ 88
3.4 Converting resistance into suction and correcting for temperature .................. 95
3.5 Field measurement of soil water suction with gypsum blocks....................... 101
3.6 Estimating soil water retention curves for the site ......................................... 104
3.7 Validation against neutron-probe soil water content...................................... 109
3.8 Conclusions .................................................................................................... 119
Chapter 4 Searching for genetic diversity in rooting depth and its relationship with
crop growth and yield ............................................................................................................ 123
4.1 Introduction .................................................................................................... 123
4.2 Materials and methods ................................................................................... 126
4.3 Results ............................................................................................................ 129
4.3.1 Differences in shoot and root traits between genotypes and environments
129
4.3.2 Genotypic differences in rooting depth, total root length, biomass and
yield 130
4.3.3 Relationships between rooting depth, total root length, biomass and yield
134
4.3.4 Plant height, root and shoot growth ........................................................ 138
4.4 Discussion ...................................................................................................... 140
4.5 Conclusions .................................................................................................... 142
Chapter 5 Relating deep-root length with water use, canopy temperature and grain yield 145
5.1 Introduction .................................................................................................... 145
5.2 Materials and methods ................................................................................... 148
xiii
5.2.1 Growing conditions ................................................................................. 148
5.2.2 Measuring canopy temperature ............................................................... 148
5.2.3 Calculating a crop water-stress index...................................................... 150
5.2.4 Calculating water use and the fraction of plant available water ............. 152
5.2.5 Measuring root length density along the soil profile ............................... 154
5.3 Results ............................................................................................................ 154
5.3.1 Growing conditions ................................................................................. 154
5.3.2 Root length density distributions............................................................. 154
5.3.3 Canopy temperature and water use ......................................................... 155
5.3.4 Crop water-stress index ........................................................................... 161
5.3.5 Crop growth and water use...................................................................... 170
5.4 Discussion....................................................................................................... 172
5.4.1 Water use was higher in Gregory than Derrimut, but differences in root
length were not demonstrable ............................................................................... 173
5.4.2 Canopy temperature was cooler in Gregory than Derrimut .................... 174
5.4.3 CWSI was a better surrogate of plant water-status than CTD ................ 175
5.4.4 Daily changes in CWSI did not correlate with water use ....................... 175
5.4.5 Trial size is more important than continuous monitoring for detecting
overall differences between genotypes—but continuous monitoring can reveal
genotype interactions with time ............................................................................ 177
5.4.6 Trial size also remains important for detecting smaller genotype by time
interactions overall ................................................................................................ 178
5.5 Conclusions .................................................................................................... 179
Chapter 6 General discussion and implications for crop improvement ............................181
6.1 Overview of the thesis .................................................................................... 181
6.2 The benefits of deeper roots to wheat growth and yield ................................ 182
6.3 Ways of improving root depth by breeding.................................................... 185
6.4 Phenotyping for deep roots............................................................................. 192
xiv
6.5 Future research ............................................................................................... 194
6.5.1 Automation of soil coring ....................................................................... 194
6.5.2 Measuring root function without measuring roots .................................. 195
6.5.3 Enhanced root vertical distribution, trade-off-free opportunities and root
self/non-self recognition ....................................................................................... 198
6.6 Conclusion...................................................................................................... 199
Appendix A ASReml-R models........................................................................................ 201
Appendix B Validation of gypsum blocks calibration against neutron probe.................. 203
References .............................................................................................................................. 209
xv
List of Figures
Figure 2.1. Location of the 6 field experiments and their corresponding sites (Creek,
Quarantine and Valley). .................................................................................................. 49
Figure 2.2. Soil chemical characteristics measured during May 2011, just before sowing
of experiments 1 and 2. Soil electrical conductivity, soil pH, soil ammonia concentration, and soil nitrate over nitrite concentration as a function of soil depth. Bars are the standard errors of the means. Colours indicate the soil of origin: Creek or Valley.
......................................................................................................................................... 52
Figure 2.3. Soil physical characteristics measured during May 2013, just before sowing
of experiment 4. Soil bulk density and air-filled porosity as a function of soil depth. Air-filled porosity was estimated from bulk density, BD, and water contents per layers at the start of the season, θ (cm3 cm-3), and an average soil particle density, PD (g cm-3) of
2.65. Then air-filled porosity - - . The dashed vertical line in the right panel indicates the theoretical limit of air-filled porosity below which most soils produce anoxia to roots (Wesseling et al. 1957). Colours indicate the soil of origin: Creek or
Valley. Bars are the standard errors of the means. .......................................................... 53
Figure 2.4. Rainfall and reference evapotranspiration (top) and maximum and minimum monthly-mean temperatures (bottom) for the four years when experiments were
conducted and a long term period from 1889 to 2014 for the Ginninderra Region. Boxplots indicate percentiles 0.10, 0.25, 0.50, 0.75, and 0.90, where points represents
outliers (i.e. smaller than percentile 0.10 or larger than percentile 0.90). In each panel, numbers on the top-right corner indicate total annual rainfall (red) and total annual evapotranspiration (blue). For the period 1889-2014 rainfall and evapotranspiration are
described as mean ± standard error (n = 125 years). Data were obtained from http://www.longpaddock.qld.gov.au/. ............................................................................. 54
Figure 2.5. Diagrams representing plots in the 6 experiments. Above-ground biomass and yield were sampled in the areas limited by dashed lines, and soil coring to sample roots was done over the same areas after sampling biomass. This was done only at
maturity in experiments 1, 5 and 6 (Zadoks 87); at anthesis (Zadoks 65) and maturity in experiment 4; and at stem elongation (Zadoks 31), anthesis and maturity in experiment
3. Plots in experiment 5 and 6 were similar to plots of experiment 4, except that they only had one biomass cut at maturity and one infrared thermometer (rather than two), and they lacked neutron probe access tubes. ................................................................... 56
Figure 2.6. Selection of genotypes at experiment 1 (2011) to be sown in experiment 3 (2012). Rooting depth at maturity for all varieties sampled in experiment 1 (left), and
above-ground biomass as a function of the sum of root counts (root tips) along the soil profile, both at maturity as well. Coloured dots are the selected genotypes. Dots are means of 4 replicates and bars are standard errors of those means. Grey dots are the rest
of the varieties sown in experiment 1, which were not selected to be sown in experiment 3. ...................................................................................................................................... 61
xvi
Figure 2.7. Selection of genotypes at 'hill plots' experiment (2011) to be sown in experiment 3 (2012). Rooting depth at maturity for the 10% deepest and the 10% shallowest roots in the 'hill plots' experiment (left), and above-ground biomass as a
function of the sum of root counts (root tips) along the soil profile, both at maturity as well. Coloured dots are the selected genotypes. Dots are means of 4 replicates and bars
are standard errors of those means. Grey dots are the rest of the varieties sown in the 'hill plots' experiment, which were not selected to be sown in experiment 3. ................ 62
Figure 2.9. Statistical design of experiment 2. The experiment was sown during 2011 in
the Valley site (yellow podzolic soil). Colour specified as 'rest' (pale green) identify plots that were not sampled, and from those, 'generic' was a wheat genotype that was
used as a buffer. .............................................................................................................. 67
Figure 2.10. Statistical design of experiment 3. The experiment was sown during 2012 in the Creek site (red podzolic soil). Colour specified as 'lodged' (pale green) identify
plots that were not sampled because they were stem-lodged at anthesis........................ 68
Figure 2.11. Statistical design of experiment 4. The experiment was sown during 2013
in the Creek site (red podzolic soil). Two commercial wheat varieties, with supposedly different root depth at maturity, were sown in combination with 2 nitrogen fertilization treatments aimed to maximise differences in root growth: 0 and 200 kg N ha-1 applied at
the stem elongation stage (Zadoks 31). However, and unfortunately, plots were mistakenly fertilised, and therefore blocks 2 and 4 don't have all the possible treatment
combinations—i.e. in block 2, Derrimut fertilised with 200 appears in two plots, and Derrimut without fertilization none. Fortunately, nitrogen had no statistical effect over shoot or root growth, so the nitrogen treatment is later disregarded. ............................. 69
Figure 2.12. Statistical design of experiment 5. The experiment was sown during 2014 in the Creek site (red podzolic soil). Four wheat genotypes where sown, but just 2 are
evaluated in this thesis: Derrimut and Gregory, same genotypes as in experiment 4. Two sowing densities were sown, a low density of 100 and a high density of 200 plants m-2. The design is a split-plot design with 4 replicates in blocks. Experiment 6 was sown
during 2014 and had the same experimental design as experiment 5, but it differed in its location: it was sown in the Quarantine site, in a yellow podzolic soil (Figure 2.1). ..... 70
Figure 2.14. Comparison of stained and non-stained samples with Toluidine blue before scanning. The dashed line is y = x and the full line is the best linear fit. It can be seen that the linear fit and the dashed line fall apart in the case of average root diameter and
root volume, but they are almost equal for total root length—therefore, staining the root samples with Toluidine blue does not improve the accuracy of measuring root length. 76
Figure 2.15. Models used to predict root length density from root count density according to soil depth. The graph shows root length density as a function of root count density as measured in the field. Soil depths are denoted by panels, with titles in cm of
soil depth. Blue lines are the non-linear fittings with equation as presented in model (2.1) and parameters and statistics presented in Table 2.7. ............................................ 78
xvii
Figure 3.1. Procedures followed to sample bulk density. For the top 60 cm layers, a hand corer was used, which allowed more cautioned sampling in those layers that are more prone to crumbling (a). We dug with a shovel between sampling layers 20, 40 and
60 cm depth, avoiding compaction of the soil to be sampled (b, c). For soil layers deeper than 60 cm, we cored with the same hydraulic corer as described for sampling roots,
where the coring tube was actually wider (6.5 cm diameter) (d). After removal from the soil, soil samples were levelled with a scraper in order to have a precise, known volume (e). ................................................................................................................................... 87
Figure 3.2. Soil water content as a function of the neutron count ratio during 2012 and 2013 for depths of 20 cm and deeper than 20 cm. The full line shows the regression line,
best fit, for both years. The dashed lines are the equations by Dr Warren Bond, detailed later in Figure 3.3. ........................................................................................................... 89
Figure 3.3. Neutron probe generic equations estimated by Dr Warren Bond. The figure
shows soil water content as a function of the neutron count ratio, the ratio between the neutrons measured in the soil and the neutrons that can be counted into a tank full of
water. The latter is used as a reference measurement to normalise neutron readings that could otherwise vary between devices. ........................................................................... 90
Figure 3.4. Validation of Dr Warren Bond's predictions of soil water content from
neutron probe data. The graph shows measured soil water content (m3 m-3) as a function of neutron count ratio (unitless) for soil beneath 20 cm at 3 locations in New South
Wales: Corowa, Ginninderra and Wagga Wagga. Notice that the parameters of the curve (slope and intercept) are closely similar to those of soil deeper than 20 cm in Figure 3.3 ......................................................................................................................................... 91
Figure 3.5. Soil water content estimated from Dr Warren Bond's equations (Figure 3.3) against soil water content calculated from gravimetric water contents at the start and end
of the season of experiments 3 and 4 and the soil bulk density measured at the end of experiment 4. These comparisons serve as validations of neutron probe predictions against direct coring data. R2 are indicated for the linear regressions between both
variables for each soil layer. ............................................................................................ 94
Figure 3.6. Gypsum block suction (kPa) as a function of the logarithm (base 10) of its
electrical resistance (Ω, Ohms). The line shows the equation , where is the logarithm base 10 of the resistance in Ohms (Johnston 2000). The inset shows the same
data in a regular axis for the electrical resistance. ........................................................... 97
Figure 3.7. Trajectories of water suction readings (up) and the pressure-chamber
temperature (down) during the calibrations. Arrows indicate the time when the readings reached a plateau (equilibrium), and hence were recorded. Numbers above those arrows (kPa) indicate the pneumatic pressure inside the chamber by that time. The colour of
lines indicates each of the 4 gypsum blocks placed inside the chamber. These sets of 4 suction readings per temperature and per pneumatic pressure recorded at equilibrium are
later plotted in Figure 3.9. ............................................................................................... 99
xviii
Figure 3.8. Idem Figure 3.7 but for a water suction of 1500 kPa. The figure was separated from Figure 3.7 to show the longer time needed for equilibrium at a greater suction. ............................................................................................................................ 99
Figure 3.9. Suction given by 4 gypsum blocks as a function of the suction in the pressure plate at three bath temperatures of 5, 15 and 25 °C. The dashed line represents
y = x. The graph on the right summarises how the slopes of the lines from the graph on the left change as a function of temperature. The latter is the equation that I used to correct suctions by temperature: y = -0.023 x + 1.472. ................................................ 100
Figure 3.10. Soil water suction, uncorrected or corrected by soil temperature (upper panel), and soil temperature (lower panel) across time for soil at 20 cm depth in
experiment 1 during 2011. ............................................................................................ 101
Figure 3.11. Installation of gypsum blocks and thermistors. The same hydraulic press mounted to a tractor that was used for soil coring when sampling roots was used for
digging holes in the centre of plots for installing gypsum blocks, albeit a thinner metallic core was used (a, b). Gypsum blocks were soaked in water the night before (d),
and covered in mud made from the same soil layer where they were supposed to be buried (c). Thermistors were deployed as well so that soil temperature could be recorded. Frame (e) shows the white cable that worked as an extension necessary to
reach the desired soil depth, joint to a thermistor (black wire) with a lead sinker at the tip. After reaching the desired soil depth, the hole where either a gypsum block or a
thermistor was installed was filled with bentonite to prevent the preferential flow of water along the walls of the hole (f). ............................................................................ 103
Figure 3.12. Example water retention curves for typical soils. Data were generated from
default water retention curves present in the software RETC (van Genuchten et al. 1991). ............................................................................................................................ 105
Figure 3.13. Water retention curves estimated for the site in experiment 4. The vertical axis is soil water content and the horizontal axis the soil water suction. Samples in the paddock were taken at two extremes of the experimental area: north and south. Soil
water contents at suctions greater than zero were measured in a pressure plate apparatus. Soil water content at complete saturation (i.e. suction equal zero) was estimated as
- as Paydar and Cresswell (1996)........................................................... 109
Figure 3.14. Soil water content predicted by the water retention curves (red lines, same curves as in Figure 3.13), and soil water content measured by neutron probe (blue dots),
as a function of the soil water suction measured by gypsum blocks. ........................... 111
Figure 3.15. A closer look at Figure 3.14, narrowing the focus within soil suctions from 20 to 60 kPa. The graph shows the disagreement between the water retention curves
measured in the pressure plate (red line) and the data measured with neutron probe at suctions less than 50 kPa (blue dots), where points cluster at higher values than what the
pressure-plate water retention predicts (20 and 40 cm depth). The blue line shows a surrogate water retention curve that results from fitting the gypsum-neutron probe data
xix
with RETC (van Genuchten et al. 1991). The latter is however unrealistic and will be later discarded, as explained in the main text. ............................................................... 113
Figure 3.16. The start of the calibration of gypsum blocks, where 'glistening point'
shows the approximate period when the soil surrounding the gypsum blocks was shining because of being saturated with water. The graph suggests that, even when fully
saturated, gypsum blocks are blind to suctions under c. 50 kPa. .................................. 114
Figure 3.17. Surrogate water retention curves obtained by fitting soil water suction measured by gypsum blocks to soil water content measured by neutron probe in the
field. Dots are neutron-probe-measured soil water content (circles are data < 50 kPa, triangles are suction > 50 kPa). Notice that the parameter was fixed as in equation
(3.4). Two possible surrogate water retention curves can be obtained. First, in blue, the lines show water retention curves fitted using all data. Second, in green, the lines show
the water retention curves fitted after excluding data < 50 kPa. Reasons for excluding data whose suctions are less than 50 kPa are detailed in the main text and supported by
Figure 3.16. ................................................................................................................... 116
Figure 4.1. Root length density distributions across experiments and genotypes. Lines represent root length densities every 10 cm, each of which is a mean of 4 replicates.
Experiments are designated as colours, as in the top-left first panel. Genotypes are detailed in the title of each panel. From those, Bogong and Currency are triticales; the
rest are wheat varieties. ................................................................................................. 130
Figure 4.2. (a) Root length, (b) rooting depth, (c) shoot biomass and (d) grain yield for each group of genotypes in experiments 1 and 2. Each bar is the predicted mean,
generated by the ASReml model, of the genotype group when combining experiments 1 and 2. Error bars are standard errors for each mean. Above and in the middle of group
bars to be compared, p-values indicate significance of their differences; where none is present comparisons are not statistically significant. .................................................... 132
Figure 4.4. Shoot biomass (left), grain yield (centre) and harvest index (right) as a
function of the root length per soil layer (rows) in each of the three experiments. Points are the predicted means, generated by the ASReml model, for each genotype in each
experiment. R2 is the adjusted R-squared and p is the p-value for the straight- line fit relating either shoot biomass or grain yield with root length deeper than 120 cm only at experiment 3. Note the change in scale in the x-axis across depth intervals. ................ 137
Figure 4.5. Predicted genotypic means, modelled from experiments 1 and 2 combined, of rooting depth (a), shoot biomass (b), and grain yield (c) as a function of final canopy
height achieved by crops after anthesis. Bars are the standard errors of the predicted means. The 8 genotypes evaluated are those differing in point mutations for the dwarfing genes (Chandler and Harding 2013). Lines represent the best fit with the
following statistics: (a) R2 = 0.47, p < 0.06; (b) R2 = 0.44, p < 0.07; (c) R2 = 0.63, p < 0.02. ............................................................................................................................... 139
xx
Figure 5.1. Pictures showing the infrared thermometers used to measure canopy temperature hour after hour. On the left, it can be better appreciated that they were positioned in a 45° angle from the vertical line. On the right, it can be seen more about
its contruction: a PVC tube inside which the electronic components are installed, and a frontal aluminium head with a hole in the middle where the infrared sensor lays. ...... 150
Figure 5.2. Graphical depiction of the concept of crop water-stress index (CWSI) developed by Idso et al. (1981) and Idso (1982). The CWSI is a ratio between how far the current CTD of a crop is from the non-water-stressed CTD, and the maximum
possible difference in CTD with the non-water-stressed CTD. The current CTD of the crop is exemplified in the graph as a black point. The non-water-stressed CTD (CTD
lower limit = CTDLL) is determined by equation (5.2) and is represented by the blue line. The maximum CTD at the current air temperature (CTD upper limit = CTDUL) is the top grey line. The segment representing the distance between CTD and CTDLL is the
numerator, and the segment between CTDUL and CTDLL the denominator, of equation (5.1). .............................................................................................................................. 152
Figure 5.3. Root length density distributed along soil depth for the varieties Derrimut and Gregory during 2013 (left) and 2014 (centre and right). The complete profile is plotted on top, and a subset, at 100 cm and deeper, is plotted at the bottom for better
appreciation of small root length density values. Points are means, and horizontal bars are standard errors, of 8 replicates. For data from 2014, an extra site is shown (right).
For that extra site only data of root length density is shown in this thesis, since air temperature and soil water characteristics were not measured, making estimations of canopy temperature deviation and soil water content inaccurate. ................................ 155
Figure 5.4. Trajectory of canopy temperature deviation from air temperature (CTD) at solar noon ± 2 hours, during 2013 and 2014 (top); water use per day during 2013 and
2014 (middle); vapour pressure deficit (VPD) at solar noon ± 2 hours; and rainfall per day during 2013 and 2014 (bottom). Rainfall bars are stacked on top of each other when coinciding between years. Anthesis and physiological maturity dates are indicated,
respectively, by a vertical, dashed line the 22nd October and another the 12th December. For CTD and water use, points are the means, and bars the standard errors, of 8
replicates. ...................................................................................................................... 156
Figure 5.5. Relative water use per soil layer and per day for Derrimut and Gregory during 2013 (top) and 2014 (middle), and rainfall per day (bottom). Water use is
relative to the total water use per day, and it is indicated by coloured areas. Numbers inside those areas are their soil depths, in cm. Rainfall bars are stacked on top of each
other when coinciding between years. Vertical, dashed lines across panels indicate anthesis (left) and physiological maturity (right) dates as in Figure 5.4. Relative water uses per soil layer are means of 8 replicates. ................................................................ 158
Figure 5.6. Accumulated water use from anthesis to maturity along the soil profile for Derrimut and Gregory during 2013 (left) and 2014 (right). Accumulated water use was
determined as the sum of every negative change in volumetric water content (m3 m-3) measured by gypsum blocks between two successive days. Points are means, and
xxi
horizontal bars are standard errors, of 8 replicates. Asterisks at soil depths 100 and 140 during 2013 indicate statistically significant differences (p < 0.05). Other soil depths during 2013 and 2014 did not differ statistically. ......................................................... 160
Figure 5.7. Soil water content along the soil profile, at the start of the season (black line and triangles), and at maturity of each genotype (coloured points and lines) measured by
neutron probe during 2013. Points are the means, and horizontal bars the standard errors, of 8 replicates in the case of soil water content at maturity. Initial soil water content is a mean of all plots in the trial, as they did not vary appreciably. ................. 161
Figure 5.8. Canopy temperature deviation around solar noon (CTD) when fitted within the framework of Idso et al. (1981) and Idso (1982) during our experiments in 2013 and
2014. Each individual point is the CTD of each of 8 replicates per genotype, either Derrimut (pale red) or Gregory (pale blue). CTDUL and CTDLL represent, respectively, the CTD upper limit and CTD lower limit as explained in Figure 5.2. ........................ 162
Figure 5.9. Crop water-stress index (CWSI) calculated from canopy temperature (top), fraction of plant available water (FPAW) (middle), and rainfall per day (bottom) across
time during 2013. Vertical, dashed lines across panels indicate anthesis (left) and physiological maturity (right) dates as in Figure 5.4. For CWSI and FPAW, points are the means, and bars the standard errors, of 8 replicates. For CWSI, data are missing
between 20th October and 7th November due to malfunction of the infrared thermometers' data loggers............................................................................................ 165
Figure 5.10. Crop water-stress index (CWSI) calculated from canopy temperature (top), fraction of plant available water (FPAW) (middle), and rainfall per day (bottom) across time during 2014. Vertical, dashed lines across panels indicate anthesis (left) and
physiological maturity (right) dates as in Figure 5.4. For CWSI and PAW, points are the means, and bars the standard errors, of 8 replicates. ..................................................... 167
Figure 5.11. A figure showing how crop water-stress index (CWSI) is a better predictor of the fraction of plant available water than the canopy temperature deviation from air temperature (CTD) alone. The figure shows either CTD (top row) or CSWI (bottom
row) as a function of the fraction of soil available water (soil water held between suctions from 30 to 1500 kPa) during 2013 (left) or 2014 (right). Each data point is the
mean of 8 replicates per genotype, for genotypes Derrimut (pale red) or Gregory (pale blue). R2 are the r-squared values fitted per genotype per year. ................................... 168
Figure 5.12. Crop water stress index (CWSI) of Gregory minus Derrimut, as a function
of water use of Gregory minus Derrimut during 2013 and 2014. Axes are the values of Gregory minus the values of Derrimut—i.e. CWSI Gregory – CWSI Derrimut, and
(water use Gregory) – (water use Derrimut). For instance, a value in the y-axis of -0.1 could mean that the CWSI of Gregory was 0.3 (less water-stressed) while the CWSI of Derrimut was 0.4 (more water-stressed). Correlation between ordinate and abscissa is
nil, but data fall within the quadrant of negative difference in CWSI corresponding to positive difference in water use between Gregory and Derrimut—more water use
corresponded with less water stress. ............................................................................. 170
xxii
Figure 6.1. Simulated water supply-demand ratio for four environment types that typify the most common water balances for wheat across the Australian wheat-belt as a function of thermal time accumulated from anthesis (base temperature = 0 °C).
Numbers indicate 4 possible environment types: environment 1: no or light stress; environment 2: post-flowering mild stress; environment 3: stresses that began before
flowering but were relieved during grain filling; and environment 4: severe stresses that began during the vegetative period and lasted until maturity. Redrawn from Chenu et al. (2013 fig. 5)................................................................................................................... 191
Figure 6.2. Depiction of the method for determining the maximum water uptake per day per soil layer (kl). On the left, the graph shows the available soil water content at a
given time t ( ) as a function of time since sowing for three soil depths. The parameters of equations (6.1) and (6.2) are indicated as well. On the right, the kl calculated as in the
graph in the left are plotted along soil depth for two sunflower hybrids. Bars are standard errors. It must be noted that the data on the left is peanut and therefore does not
relate to the data on the right; it only serves as an example. The figure was redrawn from Dardanelli et al. (1997). ....................................................................................... 197
xxiii
List of Tables
Table 2.1. The classification of the soils of each experiment site under Australian (Stace
et al. 1968; Northcote 1971; Isbell 1993) and American (Soil Survey Staff 1999) classification systems. ..................................................................................................... 50
Table 2.2. Depth, percentages of sand, silt and clay, and field texture of the horizons of both soils evaluated in this study. Data were taken from Sleeman (1979 pp. 17, 21). ... 50
Table 2.3. Agronomic practices for each experiment, including sowing date, sowing
density (pl ha-1), fertilizer applied at sowing (kg ha-1), nitrogenous fertilizer applied at the beginning of stem elongation (GS31, (Zadoks et al. 1974)) (kg N ha-1), and harvest
date. ................................................................................................................................. 57
Table 2.4. Agronomic characteristics of Derrimut and Gregory, the wheat varieties sown in Experiment 4. .............................................................................................................. 59
Table 2.5. List of genotypes, and their characteristics, sown in the three field experiments. Genotypes sown during 2011 were the same in both experiments done
during that season. ........................................................................................................... 63
Table 2.6. List of measurements done in the field in each year. Measurements were taken at specific developmental stages (root and shoot samplings), once or twice in the
whole season (bulk density, water retention parameters, water content by coring), weekly (water content by neutron moderation) or by continuous logging (water tension
with gypsum blocks and canopy temperature). Stages of crop development were stem elongation (SE, growth stage 31, Zadoks et al., 1974), anthesis (A, growth stage 65) and physiological maturity (M, growth stage 87). ................................................................. 71
Table 2.7. Parameters of the equation , describing the root
length density, (cm cm-3), as a function of the root count density, (cm-2). Each model was fitted separately for each soil layer. Significance levels of each estimate are:
0 '***' 0.001 '**' 0.01 '*' 0.05. SE is the standard error of each estimate. RMSE is the root mean square error and R2 is the coefficient of determination of each model. ......... 79
Table 4.1. Predicted genotypic means of rooting depth, root length per unit area, shoot
biomass and grain yield per experiment. Asterisks (*) or a single dot (.) indicate statistical significance, for genotype, experiment and the interaction between both, as
follows: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 'NS' 1. Values within brackets are the least significant difference, calculated as twice the maximum standard error for pairwise comparisons given by the ASReml model. ................................................................... 133
Table 5.1. Means of shoot biomass, grain yield, harvest index, and cumulative water use during grain filling, per year (top 3 rows), per genotype (middle 3 rows), and per year ×
genotype combination (bottom 5 rows). P values are specific to the rows above them. n.s. indicate comparisons that were not significant. ...................................................... 171
25
Chapter 1 Introduction: review of the potential
of root depth to increase crop yield
1.1 Introduction
Projections of global food demand for the year 2050 tell us that we must double the
amount of food that we produce today. Forecasts of whether we can or cannot achieve
this are not encouraging, given that the rates of farm yield improvement for the major
grain crops—wheat, soybean, maize, and rice—are smaller than what they should be in
order to feed the world by 2050 (Ray et al. 2013). The issue is even worse if we
consider that this large increase in food production needs to be achieved with reduced
water supply, since fresh water demanded from cities and other sinks will compete with
fresh water required by irrigation. Therefore, rather than relying on irrigation for grain
yield, increasing yield per unit of water supply in rain-fed agriculture seems a more
appropriate, sustainable alternative. But climate change is also expected to make rainfall
more uncertain in some areas and droughts are expected to be more unpredictable as
well as severe (Gornall et al. 2010). Under this scenario, it seems that the best
alternative is to increase the resilience of agricultural systems to scarce water
availability, and deeper root systems may offer a solution. Since most of the increase in
grain production must come from rain-fed agricultural systems, while considering water
as the limiting resource, we can learn some lessons from water-limited wheat
production.
26 Chapter 1
1.1.1 The framework of water-limited grain yield
We can understand water-limited grain yield by dividing the analysis between a
qualitative framework, where ways in which a crop can face water shortages are
categorised, and a quantitative framework, where grain yield is formalised as a function
of water capture by crops.
A qualitative framework to understand water-limited yield was proposed by Ludlow and
Muchow (1990). Within this framework, a crop facing drought can opt to escape or to
resist. It could escape by completing its growing cycle just before the onset of the dry
period at the beginning of summer, as is the case in Mediterranean environments.
'Escaping' was indeed one of the biggest innovations in wheat breeding in Australia,
where controlling time of flowering so that phenology matched water availability, was
one the breeding traits with the highest impact (Richards 1991). Fitting crop
development within the period with most benign water availability during the season, so
that crops can escape drought, is the primary aim when breeding for water-limited
environments.
A crop can, on the other hand, resist a drought by either tolerating or avoiding it.
Tolerating drought is not such a widespread strategy within grain crops, and as such
won't be a matter of this thesis; but as a matter of reference, it is related to protoplasmic
cell resistance to dehydration (Ludlow and Muchow 1990), so that cells can remain
alive. Avoiding drought, on the other hand, refers to avoid in the sense of eluding leaf
dehydration by keeping leaves supplied with water as needed for transpiration. It
involves the regulation of water loss by leaves, and the maintenance of water uptake,
1.1 Introduction 27
possibly by deeper roots. Some examples of dehydration avoidance involve reductions
in water loss by direct evaporation from the soil, which are possible by means of faster
leaf area expansion in early stages of the crop—as is the case of increased early vigour
in wheat (Rebetzke and Richards 1999; Rebetzke et al. 2004)—or by strategic
management, such as earlier sowing date (Kirkegaard and Hunt 2010). Another way of
controlling water loss is by modification of leaf geometry, as in the case of leaf rolling.
Leaves roll when evaporative demand is highest during the middle of the day, thereby
reducing their exposed leaf area and consequently transpiration, conserving water
(Sirault 2007). As it will be developed onwards, the core matter of this thesis is indeed
dehydration avoidance, specifically by means of deeper roots that can extract water
from deep soil layers.
The second way of looking at water-limited yield is through a quantitative framework.
This framework considers yield to be a function of the most limiting resource, water,
and is represented by the following model deduced by Passioura (1977):
(1.1)
where water used refers here to the accumulated transpiration (mm), water-use
efficiency is the amount of above-ground biomass produced per unit transpiration (kg
ha-1 mm-1), and harvest index is the ratio of grain biomass to total above-ground
biomass (unitless).
The first two terms, water used (WU) and water-use efficiency (WUE), are sometimes
ambiguous because the spatial and temporal scales on which they are based are
occasionally not defined (Passioura 2007). For instance, WUE could refer to the ratio of
28 Chapter 1
CO2 assimilation to transpiration within a square centimetre of leaf, which in regard to
time would happen almost instantly, or, on a larger scale, it can refer to the ratio of grain
yield to the amount of water input into the paddock—either as rain or irrigation, or
both—within the duration of a whole crop season. On the other hand, harvest index, a
ratio between two quantities that vary in congruence (grain mass to total above-ground
plant mass), is invariably correct—e.g. if grain mass is expressed as kg ha-1 so will be
biomass.
Now that care has been taken about the various meanings of the terms, especially the
most ambiguous term, WUE, I'll refer to them in order of increasing relevance for the
statement of this thesis in relation to root depth.
WUE is the term that, in principle, would be less affected by the rooting depth of the
crop than the other terms. This is because it depends on both extrinsic environmental
factors and intrinsic factors unrelated to root function. From the extrinsic environmental
factors, the most influential over WUE is vapour pressure deficit (VPD), a measure of
'how dry' the atmosphere is. This is because the higher the VPD the more water
molecules are traded per molecule of CO2 that is incorporated by photosynthesis.
Greater VPD, therefore, reduces WUE (Kemanian et al. 2005). From the intrinsic
factors, carbon isotope discrimination (CID), the ratio between the less abundant isotope
13C and the more common 12C in plant tissues relative to the same ratio in the
atmosphere, is inversely related to WUE at the leaf level (Condon et al. 2002). This
relationship enabled phenotyping for more water-use efficient wheat varieties, since
measuring CID in plant tissues, a sort of autopsy of the crop, is much easier than
measuring cumulative water use throughout the season and biomass at maturity.
1.1 Introduction 29
However, the problem was not that straightforward. When scaling from leaf to crop,
Condon et al. (2002) showed that CID did not always predict WUE at this higher-order
level of organization. When crops relied on stored soil moisture to grow, the CID was
inversely related to WUE, as expected, but when in-season rainfall was present, the
relationship was reversed, with CID being directly related to WUE at the crop level.
Whichever way WUE affects water-limited yield, it is certainly unrelated to root
function. There is one way, though, in which more vigorous root systems could affect
WUE at the crop level: by increasing water use early in the season. For winter cereals
like wheat, this coincides with a time in the season when VPD is low and therefore dry
matter is 'cheap' to produce in relation to its cost in water use (Gomez-Macpherson and
Richards 1995). This could be the case for crops like barley (Hordeum vulgare L.) or
triticale (× Triticosecale) which have greater leaf area expansion rates and apparently
greater root depth than wheat early in the season. Under these conditions, if barley and
triticale could exploit soil water faster than wheat early in the season by means of a
larger root system coupled with a larger leaf area, then they could grow more biomass
with a high WUE. Therefore, more vigorous roots early in crop development, either
deeper or longer, could slightly increase WUE, but in other circumstances root growth
would not affect WUE.
Harvest index (HI) can vary slightly between genotypes and can be affected by stress
conditions during the reproductive period, like drought. Changes in HI due to breeding
were responsible for the so called 'Green Revolution'. Before this revolution wheat
plants used to have heavier, longer stems. By intensive breeding in the field, crossing a
wide range of wheat cultivars, Norman Borlaug and colleagues reduced the height of the
30 Chapter 1
plant by introducing Rht dwarfing genes, and improved disease resistance. While
reducing plant height, they possibly altered the competition for assimilates between
stems in favour of grains, increasing HI and yield. If conditions for plant growth are
potential (i.e. only limited by incoming radiation and temperature; not water, nutrients
or disease), HI only depends on the genotype. On the other hand, under stressful
growing conditions HI can be drastically reduced. This reduction is more severe in
some stages of crop development than other stages, and so this gives rise to the
definition of critical periods for yield determination, when HI is most sensitive to
changes is resource availability and therefore stresses.
Knowing about the critical period of yield determination helps one to understand how
deeper roots can affect HI of wheat. The critical period for yield determination in many
annual grain crops occurs almost always around flowering, when the number of grains
that the plant will have is determined (Fischer 1985). Once the critical period has ended,
the number of grains is fixed. Because grains have a narrow range of possible sizes that
they can grow (Sadras 2007), a ceiling for achievable yield is set. Subsequent benign
conditions cannot repair the reductions in grain number. HI and yield both correlate
highly with the number of grains per unit area. After the critical period has ended,
stresses over plant growth can only reduce yield by reducing the other component of
grain yield—grain weight. How could deeper roots affect grain number and grain
weight increase and harvest index?
The mechanism by which deeper root systems could increase access to water could
happen in two ways. First, they could prevent the occurrence of intermittent stresses
around flowering that reduce grain number. Second, during the grain-filling period,
1.1 Introduction 31
avoiding terminal droughts could ensure that most of the grains set during the critical
period are properly filled. The greatest impact of deeper roots seems to be by
maintaining the source during grain filling, for there is no evidence that having deeper
roots could increase the number of grains, and therefore the sink size and HI. Setting
more grains and growing more roots around flowering could be in competition similar
to that between stem growth and grain number before the Green Revolution. It is not
clear, however, how reducing stem growth by breeding, while increasing HI, affected
the growth of roots. Miralles et al. (1997) found no differences between wheat cultivars
differing in dwarfing genes, but effects of reduced stem growth on root growth are
certainly likely to happen and might still not be properly documented. I will show later
that reducing stem height could increase not only yield but rooting depth as well, where
potentially both interact with each other—i.e. less partitioning of biomass to stems, free
assimilates to grow deeper roots, more water use, more yield.
Water used (WU), the amount of water transpired by the plant, is the single most
important factor to be affected, in principle, by more vigorous root systems with deep
roots. This is because deep roots can explore volumes of soil that would otherwise be
unexploited by plants, and thus, they can increase the supply of water to the crop,
avoiding drought. But the amount of transpiration, or WU, depends on supply and
demand. The supply is determined by the capacity of the soil and the root system to
provide water to the evaporation sites in the leaves. The demand is determined by the
atmospheric vapour pressure, radiation, wind and the total leaf area. Final transpiration
will be equal to either the supply or the demand, whichever is smaller (Cowan 1965).
To illustrate this point, if on a given sunny, hot day a wheat crop could transpire 3 mm
32 Chapter 1
(i.e. the potential evapotranspiration is 3 mm) but the root system can only provide 2
mm, then transpiration will be 2 mm. On a different day, with cool and cloudy
conditions and consequently low VPD, a wheat crop with an extensive root system
growing in a wet soil that can provide 4 mm/day but with a low potential
evapotranspiration of 1 mm/day, will transpire 1 mm/day. From this view, having deep
roots does not guarantee higher WU—it only helps in providing water when it is
actually demanded. Deeper root systems can only affect the supply.
Finally, and importantly, the terms in equation (1.1) are not independent between each
other, and the interaction is most likely to happen between terms where deep roots can
have an effect: water used and harvest index. More precisely, when regarding WU, the
amount is only part of the story—the timing of the uptake is important as well. A crop
could maximize WU by depleting soil moisture before flowering, and if there is no
rainfall thereafter, there will not be enough water for grain filling. Under these
circumstances WU will be high but HI will be very low or nil, and so will be yield. This
is exemplified by Passioura (1976), who achieved higher grain yields in potted wheat
when the same amount of water was administered in equal amounts along the plants'
development than when applied all at once at the start of their life cycle. This occurred
because photosynthesis was sustained during grain filling and the plant tissues
necessary for an adequate remobilization of assimilates to the developing grains also
survived (Passioura 1976). Indeed, HI can be interpreted as a function of the fraction of
WU after anthesis over WU for the whole cropping season (Sadras and Connor 1991).
An example of this concept in expressed in field conditions is the study of Siddique et
al. (1990), where 10 wheat cultivars representing different years of release were
1.1 Introduction 33
evaluated. They found that modern cultivars used less water but more of that water was
used during the post-anthesis phase than the pre-anthesis phase. These cultivars also
yielded more through a higher harvest index. Thus, timing is as important as the
amount. Deep roots could increase the danger to crops by exhausting all soil moisture
before anthesis.
1.1.2 How can deeper root systems improve water-limited yield?
Having deeper roots can, at the same time, (i) increase the total amount of potential WU
as well as (ii) increase the risk of reducing HI so dramatically as to achieve a point
where there is no yield. Having the right balance between pre- and post-anthesis water
use is a key to maximize yield, and using as much water as possible is a key to
maximize biomass. There are therefore two opposite strategies on how a crop might
deal with water shortages. A conservative strategy minimizes WU pre-anthesis,
increasing the chances of having water left in the soil during grain filling. The opposite
strategy is where crops use as much water as possible, in a luxurious way, so that
biomass is maximized. The issue is how deep roots may affect these strategies.
1.1.2.1 Deep roots, conservative and luxurious use of water
Being conservative can increase yield by increasing HI and WUE when crops grow
primarily on stored soil moisture. This is why, as discussed before, Condon et al. (2002)
obtained higher yields when increasing WUE through less CID only in years when
wheat grew with limited water supply during grain filling. Yield benefits from being
conservative as a crop in terms of water use is also supported in a simulation modelling
34 Chapter 1
study by Sinclair et al. (2005). When sorghum limited its transpiration rate to a
maximum by closing stomata during midday, yields were increased on average by 7%,
and even higher yield increases were obtained in the worst (driest) seasons. Regarding
roots and conservative use of water, there was a successful breeding programme based
on a plant physiological framework developed by Richards and Passioura (1989).
Wheat lines were developed with reduced xylem diameter in the seminal roots. This
resulted in a larger resistance to axial flow, so plants withdrew water from the soil at a
much slower rate, leaving more water for later during the critical period of yield
determination. This increased yield when plants were growing mostly on stored soil
moisture. This is one example of conservative behaviour of water use. Are there
examples of the opposite, that is, luxurious use of water?
In environments were the soil profile could be replenished by later rainfall, crops
usually leave water unused at the bottom layers of their rooting depth (Jordan and
Miller 1980). Therefore, a crop with an 'luxurious' strategy that can exploit that unused
deep water could be more effective in the use of water. WU could be increased and
consequently yield. The advantage of being luxurious as a crop in using water is
supported by Blum (2009), who proposed the concept of efficient use of water, by
which he states that water use and harvest index are the drivers of yield in water-limited
conditions, not transpiration efficiency. He suggested that if we call water use and
harvest index 'drivers', transpiration efficiency should be called 'a passenger'. He
emphasized the value of deep root systems for maintaining transpiration during grain
filling. For Blum (2009), being water-use efficient is not the solution for breeding crops
that produce more under water-limited conditions, but rather to have crops that are
1.1 Introduction 35
better 'exploiters' of deep water. The key to increase water-limited yield, for him, was to
increase water use. He also argued that the success of Condon et al. (2002) in selecting
for lower CID related to high WUE and resulting in high yields was because in this
region of Australia, NSW, wheat relies on stored water, supporting the deduction of
Passioura (1976) of the timing of water use.
A strong argument against Blum (2009), however, is that breeding has not yet improved
the amount of total WU but rather the WUE in an evapotranspiration basis, meaning
that either transpiration efficiency was increased or direct soil evaporation was reduced.
This is shown by Campos et al. (2004) for maize in the US and Sadras and Lawson
(2013) for wheat in South Australia. Campos et al. (2004) found that new maize hybrids
used less water but produced more yield, and, interestingly, water use at shallow soil
layers was less than in old hybrids, with a slight increase in deep water use (Fig. 3 in
Campos et al., 2004). This may be telling us that total water use is a difficult trait to
improve and that it wasn't improved yet, but it certainly does not mean that it is
impossible to change. Moreover, it could be possible to develop a crop that can at the
same time exploit the deeper soil by means of deeper roots, and have a reduced stomatal
conductance during critical times of the day, around noon, when fixing CO2 is relatively
expensive in terms of water use. In fact, the analysis of Sinclair et al. (2005) is about
limiting transpiration rate to a maximum by limiting leaves conductance. Leaves
conductance is an independent trait from deep water use, and opens the possibility,
provided the traits are independent and can be separated, of breeding for both traits
pyramided in the same genotype (Wilson et al. 2015). There seems to be no reason
against having both traits in the same plant.
36 Chapter 1
1.1.2.2 Objections, and their solutions, to higher production with
deeper roots
While it is intuitive to suppose that more root length deep in the soil will increase WU
and improve the fate of crops facing droughts, there are many objections. Some
objections are related to the fact that growing more roots implies diverting assimilates to
them, assimilates that could otherwise be directly used in growing more grain. Others,
to the fact that sometimes roots, while present in deep soil with apparently sufficient
length-density and time, fail to extract the available water leaving after maturity
valuable water deep in the soil.
Investing more biomass in deep roots, because of the law of conservation of matter,
means that there would be less assimilates to invest in grains. Therefore, it could be the
case that the extra roots do not recover, in terms of grains, the amount of assimilates for
growth and maintenance respiration that were initially invested into them. This is
aggravated by the fact that roots often respire, per unit dry weight, more than the shoot
tissues (Passioura 1983). And investing in deeper roots can also compete with the
accumulation of water soluble carbohydrates (Lopes and Reynolds 2010), which often
sustain grain growth when conditions of drought prevail during grain filling. Perhaps it
could be more advantageous for grain growth just to allocate carbon to grains or to store
sugars into stems rather than investing in more root growth.
Another objection for crops with deeper roots is that there are many cases where water-
stressed crops fail to extract much of the available water in the subsoil even if sufficient
root length was present (Jordan and Miller 1980; Kirkegaard et al. 2007). Several
1.1 Introduction 37
reasons were proposed. One of those reasons is that sometimes roots do not actually
distribute evenly in the subsoil, but they rather follow the structure of the soil. For
instance, in highly structured soils, roots usually grow along the paths of less resistance,
which drives them to clump together inside macro-pores left behind by past roots or
worms or planes formed by cracks in the soil (White and Kirkegaard 2010). This soil
structure limits the effectiveness of root length density because instead of root length
being distributed evenly through the soil, it is confined to macro-pores which are more
scarcely distributed in the soil volume. Water uptake under these conditions is not
determined by root length density, but rather, by macro-pore distribution. Deery (2008)
and Deery et al. (2012) explored the possibility of this limitation being caused by the
soil or by the rhizosphere itself. Because the parameter that determined soil water flow
in the soil, soil water diffusivity, was much larger than they expected, and root
distribution was even, they concluded that the limitation was more likely within the
rhizosphere. This limitation in the rhizosphere can be caused, as they stated, by
shrinkage of the roots as the transpiration rate increased—reducing contact between
roots and soil and thereby limiting flow—or the possibility that when transpiration was
high the plant itself was increasing the resistance in the roots by means of aquaporins.
Control by the plant was suggested through aquaporins (molecular channels across
cellular membranes that are responsible for water transport in and out the cells), or
control by the root by altering the composition of exudates (Ghezzehei et al. 2015).
Moreover, a limit to water extraction by roots can be imposed by solute accumulation in
the root-soil interface, as suggested by Stirzaker and Passioura (1996).
38 Chapter 1
In spite of these objections to deep roots, there are possible solutions that could justify
breeding for them. To address the matter of the waste of carbon when investing in deep
roots rather than grains, having deeper roots does not necessarily mean that is necessary
to grow more root biomass. In fact, crops may be producing much more root length
density in the topsoil than what is actually necessary for water uptake, so why not to
shift the allocation of growth from shallow roots to deep roots? According to Deery
(2008) the optimum root length density for water uptake is about 0.1 cm cm-3. Root
length density at a soil depth of 20 cm is usually 4 cm cm-3. It is worth to introduce here
the distinction of the two types of roots that form a grass root system: seminal roots,
which originate from the seed embryo, and nodal roots, which originate from
subsequent nodes either from seed or stem tissues. Deep root length seems to be more
closely related to the growth of seminal roots. In a field study evaluating wheat, barley
and triticale, Watt et al. (2008) found that most roots present beyond 1 m depth were
first- or second-order branches of seminal roots, never nodal roots. In fact, nodal roots
have features which make them less suitable for deep water uptake: they emerge later
which gives them less time to reach deep soil layers, show less gravitropism, and are
more vulnerable to nutrient or water stress (Tennant 1976). However, roots present deep
in the soil were neither strictly seminal axes. Based on the number of xylem tracheary
vessels, Watt et al. (2008) found that these first- or second order roots differed in their
inner anatomy to the original seminal root axes. They had indeed less cross-sectional
area of xylem trachery vessels (less number and smaller vessels) that make them one to
two orders of magnitude less capable of axial water flow when compared to seminal
roots. Therefore, it seems that the capture of deep water in the soil relies on branches
1.1 Introduction 39
from seminal roots, but there is still potential to increase water uptake if seminal root
axes, rather than its branches, can reach deep soil layers. Nonetheless, whether deep
roots are seminal or nodal, the fact that root length in shallow soil is above the threshold
required to maximise water uptake indicates that there is still room to reallocate root
growth from shallow to deeper soil layers. Indeed, this shift might have been
responsible for the increase in maize yield due to breeding in United States corn belt
(Hammer et al. 2009).
There may be opportunities for breeding wheat with root growth shifted towards deep
soil layers since natural selection, while maximising individual competitiveness, may
have rejected a shift of growth to deep roots (Denison 2012, 2015). It happens that
natural selection operates much more over individual plants than over plant populations
(Denison 2012). Individual plants, in competition with neighbouring plants, can only
maximise individual performance by selfishly using resources. Therefore, an
overproduction of roots in shallow soil may be a strategy to outcompete neighbouring
plants. Moreover, it seems clear that plants can recognise self from non-self roots
(Gersani et al. 2001; Falik et al. 2003), and the response to non-self root encounters is
overproduction (Gruntman and Novoplansky 2004). This strategy of outcompeting
neighbours in a crop situation is undesirable. This relates to the pivotal research by
Donald (1968) and his definition of plant ideotypes, that is, plants with traits that make
them better allies when sown in a crop. Ideotypes have in common that while reducing
their competitive ability to a minimum the overall collective performance of the crop,
with performance defined, for instance, as their capacity to produce grain yield, is
higher than common, 'selfish' plants. In relation to the distribution of roots, a wheat
40 Chapter 1
ideotype would have more root growth allocated to deep soil and less in shallow soil
layers, and therefore competitiveness between neighbour plants is reduced and that
growth surplus can be allocated to other organs, like deep roots (Zhang et al. 1999;
Lynch 2013). Therefore, there may be opportunities for shifting root growth to deep soil
layers while reducing growth in the topsoil.
Another objection to increased investment into deeper roots is the fact that sometimes
roots are present but unable to use residual water. Evidence that breeding has improved
the WU is still lacking but there are many good reasons that it is worth the challenge.
Increasing the availability of water during the critical periods, while maintaining a
proper balance of pre- and post-anthesis WU, should increase the stability of yield.
Opportunities may also exist to breed crops with steep root angle and higher allocation
to deep rather than shallow roots (Lynch, 2013). This will serve as a justification for the
next section, where one starts wondering how deep roots could be increased:
management and most importantly, as is the focus of my thesis—breeding.
1.1.3 Improvement of root depth through management and genotype
selection
As discussed in the previous section, deep roots can increase yield in grain crops, but
how do we improve deep roots? Two alternatives seem to be available: by management
or by breeding. Management can increase deep-root length by either lengthening the
period of root downward growth or by removing soil factors that are detrimental to root
growth. An example of root-growth lengthening is sowing earlier. Roots seem to grow
mostly during the vegetative period in annual plants. When plants like wheat flower, the
1.1 Introduction 41
assimilates are partitioned to the developing grains. Organs like leaves even senesce and
in doing so remobilize their nitrogen content to satisfy the grains' need for protein. In
congruence, roots stop growing—or at least this is commonly believed (Lynch 1995). If
sowing earlier can make the period from sowing to flowering longer, and root-growth
rates are kept unchanged, then rooting depth and length can be increased (Kirkegaard
and Lilley 2007). Sowing earlier was documented by Kirkegaard and Hunt (2010),
where they found higher water-use efficiency and more water use than normal sowing
times, possibly by the contribution to total water use from deeper roots. Improving the
hospitability of soils to detrimental factors, such as acidity and its concomitant
aluminium increased concentrations and toxicity (Delhaize et al. 2004), can also
increase rooting depth and functional root length deep in the soil. Also the introduction
of break crops into rotations—in contrast with continuous wheat—has removed some
root diseases and improved soil hospitability as well (Kirkegaard and Hunt 2010).
The second alternative, breeding cereals with deeper roots, is the core of this thesis.
Relying in breeding rather than management, has the advantage that the improvements
can be cheaper and, if the traits are constitutive rather than adaptive, the benefits can be
ubiquitous. As an example, an experimental wheat variety developed to expand its leaf
area faster so that the soil can be covered quicker also had more vigorous root growth
and was better able to cope with soil pathogens (Watt et al. 2005). Another breeding
aim was to select wheat genotypes with a narrower angle of root growth so that more
root length was directed towards deeper soil layers (Manschadi et al. 2008). A
simulation model predicted that this would increase yield under water-limited
conditions (Manschadi et al. 2006). Breeding might have been already successful in
42 Chapter 1
improving deep water use in wheat, as new wheat varieties extract more water from
deep soil layers than old wheat varieties (Pask and Reynolds 2013).
If breeding is the chosen alternative to improve deep roots, how do we select for
genotypes with deeper roots? Several avenues, with different capacities of success, are
proposed. First, selection based on measuring roots activity without actually measuring
the roots, an alternative that can be illustrated in a quote by the pioneer crop modellers
De Wit and Penning de Vries (De Wit and Penning de Vries 1983): "modelling
hormone action without modelling the hormones" (Hammer et al. 2004). Dardanelli et
al. (1997) used this technique when growing sunflower in the field. They did not
measure root length density but they inferred root presence by quantifying the changes
in soil water content by using a model originally proposed by Passioura (1983). They
found genotypic differences in the maximum rate of water uptake between two
sunflower hybrids, and this example remains one of the few where genotypes differed
for the rate of water extraction in field conditions. In another study, Dardanelli et al.
(2004) applied the same simple model to predict water use by several crop species, such
as maize, soybean, sunflower, and peanut, with a high degree of accuracy. A drawback
of this technique, however, is that high water extraction could be either due to a more
effective deeper root system as to a higher stomatal conductance. Separating
conductance from root length is the key to work under this frame work. Another
drawback is that to measure the maximum uptake rates the crops have to be kept under
conditions where the demand of water from the atmosphere is higher than the
achievable supply by the roots. This can only be achieved by preventing input from
rainfall, either by sowing in arid environments or by building effective rain-out shelters.
1.1 Introduction 43
A second alternative is to anticipate rooting depth in the field by previously screening
genotypes in the glasshouse. For example, root angle is a trait that can be measured in
pots, and can be related to rooting depth of adult plants in the field (Manschadi et al.
2008). Root angle may be a stable trait whose selection at young developmental stages
of wheat can transcend towards adult stages, like anthesis. Measuring root length at
seedling stages, however, may be meaningless for wheat production in the field, for
Watt et al. (2013) found that root length of juvenile plants growing inside wet paper
rolls did not correlate with the root length of adult plants growing in the field. In other
words, differences in root length evident during early stages of plant development dilute
towards adult stages that are critical for yield determination. Therefore, measuring root
angle in young wheat plants is a promising trait but root length is not.
A third selection alternative is to directly sample roots in the field by soil coring. This is
time consuming but some automation of the coring process could increase throughput.
Digitization of the root sampling images for later analysis could further help to quantify
root presence or even root length (Wasson et al. 2012, 2014).
A fourth selection alternative is canopy temperature, a trait that can be used as a
surrogate of transpiration. Cool canopies during grain filling might indicate access to
deep water—hence deeper roots. Reynolds et al. (2007) showed that cooler canopies
during grain filling left less unused water between 30 and 120 cm soil depth. Lopes and
Reynolds (2010) linked drought adaptation to deeper root mass distribution which was
detected by cooler canopy temperature. However, caution must be taken when
interpreting canopy temperature measurements, as variables such as leaf porosity, plant
height, biomass and yield can confound the signal (Rebetzke et al. 2012).
44 Chapter 1
This thesis will focus on how to select for wheat crops with deeper roots. First, genetic
variability in rooting depth under field conditions was evaluated in a large germplasm
collection possessing diversity in stem growth and shoot vigour. Second, canopy
temperature was evaluated as a surrogate for transpiration and as a trait to select for
deeper roots under field conditions.
1.2 Thesis aims
The are two main aims of this thesis: (i) to identify genetic variability in root depth
under field conditions, in relation to biomass and yield of wheat and triticale, and (ii) to
see if measuring canopy temperature and soil water content continuously can serve as
surrogates of deep root length that is useful for field phenotyping.
1.3 Thesis outline
In Chapter 2 the many methods used to measure roots, biomass, and yield in the field
will be detailed.
Gypsum blocks were used to quantify soil water suction (kPa) in the field. The
complexities of the methods involved in this calibration to eventually determine
volumetric water content (m3 m-3) are presented in Chapter 3. Gypsum blocks enabled
soil water content to be monitored continuously in 2 wheat varieties.
Rooting depth and root length are explored under field conditions where 34 wheats and
2 triticale varieties were phenotyped. The results of these experiments are presented in
1.3 Thesis outline 45
Chapter 4, and the relationships between genetic variability in deep-root length, crop
growth, total above-ground biomass and grain yield are analysed.
In Chapter 5, the issue of whether deep-rooted wheats can be selected without digging
the ground is addressed. For this purpose, canopy temperature was used as a surrogate
for transpiration. Two commercially available high yielding wheat varieties for
Australia were investigated. Soil water, as suction (kPa) and as volumetric content (m3
m-3), was continuously monitored at different soil depths and changes compared against
canopy temperature.
In a final discussion that forms Chapter 6, the major findings of the whole thesis are
presented in relation to a unifying hypothesis.
47
Chapter 2 Field experiments used for wheat
phenotyping
2.1 Introduction
In this Chapter I'll explain many of the methods and measurements done in the field. It
is a methodological chapter, and thus, unlike the rest of the chapters, no hypothesis is
tested.
Most of the experiments were carried out in the field in common growing conditions for
wheat farming in New South Wales. A main difference was in the genotypes sown, as
we utilised lines recently developed by research in CSIRO in addition to commercially
available material. Examples of those developments are the "overgrowth" varieties by
Dr Peter Chandler (see Chandler and Harding, 2013), the high initial vigour varieties
developed by Dr Greg Rebetzke (Rebetzke et al. 2004) and the tillering inhibition
isogenic lines by Dr Richard Richards (Duggan et al., 2005). Other agronomic
conditions, exclusive to our experiments but unusual to farmers in the region, were the
fertilization treatments in experiment 4, where rates of nitrogen applied were too low
(none) or excessively high (200 kg ha-1); they were, however, appropriate for our aims
of exploring contrasting growing environments for shoots (particularly leaf area) and
roots.
In the following sections I'll explain how we performed the experiments in the field
during three consecutive seasons from 2011 to 2013, as well as how we measured roots
48 Chapter 2
and soil characteristics in the laboratory when needed. Several devices that quantified
soil water suction and content, leaf and soil temperatures, and local weather variables
for the experimental site were provided by the High Resolution Plant Phenomics Centre,
and will be detailed as well.
All field experiments were done in the Ginninderra Experimental Station, which
belongs to CSIRO and is located in the Australian Capital Territory (35.2 S, 149.08 E).
The site has a greater annual rainfall than surrounding wheat growing areas in NSW,
and thereby crops grew closer to potential conditions—that is, closer to conditions when
growth and development are only limited by radiation and temperature.
2.2 Growing conditions: soils and weather
We performed six field experiments in four consecutive growing seasons in 2011, 2012,
2013 and 2014. They were conducted in the Ginninderra Experiment Station in three
close locations but belonging to two different soil classes (Figure 2.1). Experiments 1
and 2 were done during 2011, experiment 3 during 2012, experiment 4 during 2013, and
experiments 5 and 6 during 2014.
2.2 Growing conditions: soils and weather 49
Figure 2.1. Location of the 6 field experiments and their corresponding sites (Creek, Quarantine and Valley).
Experiments 1, 3, 4 and 5 were sown in a in a Red podzolic soil (Sleeman 1979), on a
site that onwards will be called "Creek". Experiments 2 and 6 were sown in a Yellow
podzolic soil (Sleeman 1979), in the sites designated as "Valley" and ―Quarantine‖,
respectively. The classification of each soil according to Australian (Stace et al. 1968;
Northcote 1971; Isbell 1993) and American systems (Soil Survey Staff 1999) is listed in
Table 2.1. The characteristics of the horizons of each soil are described in Table 2.2.
The Red podzolic soil where experiments 1, 3, 4 and 5 were sown, allowed plants to
grow better than the Yellow podzolic soil, where experiments 2 and 6 were sown. To
50 Chapter 2
characterize the soil we performed coring until a depth of 2 m; these coring procedures
were almost identical to the root coring procedures which will be described later in
section 2.6.2, but with a larger diameter coring tube.
Table 2.1. The classification of the soils of each experiment site under Australian (Stace et al. 1968; Northcote 1971; Isbell 1993) and American (Soil Survey Staff 1999) classification systems.
Site Experiments Stace et al.
(1968)
Northcote (1971) Isbell
(1993)
Soil Survey Staff
(1999)
Creek 1, 3, 4, 5 Red podzolic soil Dy3.82 Hydrosol Oxic Paleustalf
Quarantine and Valley 2, 6 Yellow podzolic
soil
Dy3.42 Hydrosol Udic Paleustalf
Table 2.2. Depth, percentages of sand, silt and clay, and field texture of the horizons of both soils evaluated in this study. Data were taken from Sleeman (1979 pp. 17, 21).
Soil Horizon Depth (cm)a Sand (% ) Silt (% ) Clay (% ) Field texture
Red podzolic soil A1 0-8 68 20 14 Fine sandy loam
A21 8-16 69 17 13 Fine sandy loam
A22 16-25 60 19 20 Fine sandy loam
B1 25-34 29 9 57 Medium clay
B2 34-52 14 9 70 Medium clay
52-69 20 12 64 Heavy clay
69-83 24 14 57 Heavy clay
Yellow podzolic soil A1 0-5 72 18 9 Loamy sand
A21 5-18 71 21 7 Loamy sand
A22 18-38 74 11 16 Sandy loam
B21 38-48 51 16 31 Light clay
48-78 56 18 25 Clay loam
78-100 59 18 22 Clay loam
aSoil depths were registered by Sleeman (1979) until 1 m, but the soil beneath 1 m was
homogenous in both profiles and can therefore be described by the last horizon in the table.
2.2 Growing conditions: soils and weather 51
Soil chemical (Figure 2.2) and physical properties (Figure 2.3) differed between both
Creek and Valley soils. However, physical properties are the most likely candidate to
explain the reduction in shoot and root growth in that the Valley soil. A key component
of the soil physical properties of a soil is its bulk density, that is, the ratio between the
dry weight and the unaltered volume of a soil. Bulk density can be reduced by
compaction and that compromises water flow and gas exchange. It can be seen in Figure
2.3 that bulk density was always greater for the soil at the Valley at soil depths beyond
40 cm. The pore space left for gas exchange between the roots and the atmosphere, the
air-filled porosity, a function of bulk density, becomes too small, and oxygen supply
can be so slow as to limit root growth (Wesseling et al. 1957). Anoxia could then be the
reason for the impoverished growing conditions for plants, and their roots in particular,
in the Valley soil.
52 Chapter 2
Figure 2.2. Soil chemical characteristics measured during May 2011, just before sowing of experiments 1 and 2. Soil electrical conductivity, soil pH, soil ammonia concentration, and soil nitrate over nitrite concentration as a function of soil depth. Bars are the standard errors of the means. Colours indicate the soil of origin: Creek or Valley.
2.2 Growing conditions: soils and weather 53
Figure 2.3. Soil physical characteristics measured during May 2013, just before sowing of experiment 4. Soil bulk density and air-filled porosity as a function of soil depth. Air-filled porosity was estimated from bulk density, BD, and water contents per layers at the start of the season, θ (cm
3 cm
-3), and an average soil particle density, PD (g cm
-3) of 2.65. Then air-filled
porosity
. The dashed vertical line in the right panel indicates the theoretical limit
of air-filled porosity below which most soils produce anoxia to roots (Wesseling et al. 1957). Colours indicate the soil of origin: Creek or Valley. Bars are the standard errors of the means.
The weather conditions, when categorised within the climate classification system of
Köppen (1918) (see also Wikipedia, 2014), were those of a temperate class climate,
which are characterised by a lack of dry season and a warm summer (designated Cfb by
Köppen, 1918). Rainfall, reference evapotranspiration as calculated by Penman-
Monteith FAO (Zotarelli et al. 2010), minimum and maximum air temperatures for the
site were obtained from http://www.longpaddock.qld.gov.au/. Regarding these data,
rainfall varied much more than temperatures (Figure 2.4). In particular, experiment 3,
during 2012, might have started with more initial soil-water stored given the high
rainfall accumulated during March (Figure 2.4). Season 2012 was probably better in
54 Chapter 2
terms of resource availability for growth than 2011 (and this will be reflected later in
Chapter 4).
Figure 2.4. Rainfall and reference evapotranspiration (top) and maximum and minimum monthly-mean temperatures (bottom) for the four years when experiments were conducted and a long term period from 1889 to 2014 for the Ginninderra Region. Boxplots indicate percentiles 0.10, 0.25, 0.50, 0.75, and 0.90, where points represents outliers (i.e. smaller than percentile 0.10 or larger than percentile 0.90). In each panel, numbers on the top-right corner indicate total annual rainfall (red) and total annual evapotranspiration (blue). For the period 1889-2014 rainfall and evapotranspiration are described as mean ± standard error (n = 125 years). Data were obtained from http://www.longpaddock.qld.gov.au/.
2.3 Sowing and agronomic management
Crops were sown in the field in plots of varied length but constant width (1.8 m, the
width of a sowing machine, Figure 2.5). We chose different plot lengths in different
seasons as a solution between the total area of paddock available, the required number
of measurements for the season, and the number of treatments involved. As a result,
plots were 6 m long in experiments 1 and 2, 7 m in experiment 3, and 8 m in experiment
4 and 5 (Figure 2.5). All plots were sown with 10 rows distanced 0.18 m from each
other (which is a conventional row spacing for wheat growing in this region). Plots were
2.3 Sowing and agronomic management 55
fertilized, in all experiments at sowing, with a mix of 15-12-0-12 Croplift 15®
(percentages of N, P, S and K, respectively) at a rate of 100 kg ha-1. Additionally,
experiment 4 included a fertilization treatment with nitrogen, where, besides the default
fertilization with the 15-12-0-12 mix, 0 or 200 kg N ha-1 were added just before the
stem elongation growth stage (growth stage 31, Zadoks et al., 1974). Further
information about agronomic practices for each experiment are included in . Other
abiotic (nutrients) or biotic (weeds, diseases) stresses were avoided by sound
management, since we aimed to achieve growing conditions where only radiation,
temperature and water limited growth of the crops.
56 Chapter 2
Figure 2.5. Diagrams representing plots in the 6 experiments. Above-ground biomass and yield were sampled in the areas limited by dashed lines, and soil coring to sample roots was done over the same areas after sampling biomass. This was done only at maturity in experiments 1, 5 and 6 (Zadoks 87); at anthesis (Zadoks 65) and maturity in experiment 4; and at stem elongation (Zadoks 31), anthesis and maturity in experiment 3. Plots in experiment 5 and 6 were similar to plots of experiment 4, except that they only had one biomass cut at maturity and one infrared thermometer (rather than two), and they lacked neutron probe access tubes.
2.4 Germplasm sown 57
Table 2.3. Agronomic practices for each experiment, including sowing date, sowing density (pl ha
-1), fertilizer applied at sowing (kg ha
-1), nitrogenous fertilizer applied at the beginning of
stem elongation (GS31, (Zadoks et al. 1974)) (kg N ha-1
), and harvest date.
Experiment Year Site Sowing
date
Sowing
density
(pl m-2
)
Fertiliser
at sowing
(kg ha-1
)
N fertiliser
at GS31c
(kg N ha-1
)d
Harvest
date
1 2011 Creek 29th
May 150 15-12-0-12b 100 30
th Dec
2 2011 Valley 29th
May 150 15-12-0-12 100 30th
Dec
3 2012 Creek 29th
May 150 15-12-0-12 100 28th
Dec
4 2013 Creek 5th
June 150 15-12-0-12 0, 200e 23
th Dec
5 2014 Creek 2nd
June 100, 200a 15-12-0-12 100 23
th Dec
6 2014 Quarantine 2nd
June 100, 200 15-12-0-12 100 23th
Dec aExperiments 5 and 6 included a plant density treatment with levels 100 and 200 pl m
-2 (Figure
2.12). bNumbers indicate, in this order, kilograms per hectare of nitrogen, phosphorous, sulphur and
potassium. In all cases they were added as a single fertilizer mix (Croplift 15®) at a rate of 100 kg ha
-1.
cGS31 = growth stage 31, beginning of stem elongation (Zadoks et al. 1974).
dIn this case, the fertiliser used was granular urea.
eExperiment 4 included a nitrogen treatment with levels 0 and 200 kg N ha
-1 added at GS31
(Figure 2.11).
2.4 Germplasm sown
Germplasm included a wide variety of wheat (Triticum aestivum L.) and triticale (×
Triticosecale) varieties, either commercially available or experimentally developed (i.e.
not available commercially), to explore contrasting patterns of below- and above-
ground crop growth (Table 2.5). Patterns of shoot and root growth were thought to
change in response to several putative mechanisms that were tested before in laboratory
conditions. These mechanisms were (1) "overgrowth" wheat mutants, (2) inhibited
tillering, (3) increased early vigour in wheat, (4) and increased early vigour by means of
another species, i.e. triticale.
58 Chapter 2
Experiments 1 and 2 included the broadest selection of genotypes, where 34 wheat and
2 triticale varieties were sown. They were selected because they included alternative
mechanisms to enhance root growth.
Experiment 3 involved a subset of 12 wheat and triticale varieties from the previous,
original 36, but only 8 will be analysed, for the other 4 varieties were lodged by wind at
anthesis and their data are suspiciously altered. The remaining 8 varieties were selected
after analysis of experiment 1 and 2 data, because they showed to be deep rooted.
Experiment 4 was intended to explore wider ranges of rooting depth and water use by
using two contrasting fertilization treatments with nitrogen, and two commercial wheat
varieties that, after experiments performed by Dr James Hunt, at CSIRO, shown to have
contrasting vertical patterns of water use notwithstanding similar time to flowering and
similar grain yield. The fertilization treatments aimed to alter leaf area development and
presumably root growth, for it is known of the large effect that nitrogen has over leaf
expansion. The varieties were Derrimut, a variety developed for NSW, where wheat
usually thrives by using on-season water from rain falls, and Gregory, a wheat variety
developed for Queensland, where wheat yield depends more strongly on the use of
stored soil water. Their contrasting origin in terms of breeding areas implies that
Gregory may be better suited to deep water uptake than Derrimut. James Hunt observed
that when comparing both varieties in the field Gregory was able to extract more water
deeper than Derrimut, and total water used by the former was greater. Although similar
yields and greater water used indicate that Gregory is less water-use-efficient than
Derrimut, it is interesting to investigate how the former can capture more water at depth
than the latter. Moreover, mechanisms that promote greater water use could be different
2.4 Germplasm sown 59
from those that confer greater water-use efficiency. In that case, where there is no trade-
off, one could breed for varieties that have large water use and optimal water-use
efficiency. Further characteristics of these two wheat varieties are shown in Table 2.4
Table 2.4. Agronomic characteristics of Derrimut and Gregory, the wheat varieties sown in Experiment 4.
Variety Release Origin Qualityc Maturity type Morphology Height
(cm)
Derrimut 2007 Nugraina Australian Hard Medium early Semi-dwarf 80
Gregory 2004 EGAb Australian Hard Medium late Moderate height 100
aNugrain Pty Ltd, http://nuseed.com/au/
bGregory was bred by Pacific Seeds (https://www.pacificseeds.com.au/) and the Queensland
Department of Primary Industries within the Enterprise Grains Australia (EGA) joint venture. cThis is the maximum attainable quality grade (http://wheatquality.com.au/classification/how-it-
works/classes/). The actual grain quality would vary among environments.
Genotypes sown in experiment 3 were selected from experiment 1 and also from a field
trial performed by Dr Anton Wasson, from CSIRO as well. Anton's experiment was
done during 2011 at the same location as experiment 1 in the current thesis, but instead
of being sown in 6-m-long plots, it was sown in "hill plots". "Hill plots" are not groups
of equally spaced plants like those sharing a canopy in normal agronomic conditions.
Instead, they are sets of ~30 plants growing crowded in an area not bigger than 100 cm2.
They may not reflect agronomic conditions but they are certainly appropriate for
breeding purposes, where hundreds of genotypes are evaluated at one. They are sown by
hand by pouring 30 seeds down a 40 mm diameter PVC tube. For selecting genotypes
from experiment 1, to be sown in experiment 3, we selected the 8 of the deepest-rooted
varieties (Figure 2.6 left), which were as well some of the varieties producing the
60 Chapter 2
highest biomass (Figure 2.6 right). For selecting genotypes from Anton's hill plots, we
chose those that fell within the 10% deepest or the 10% shallowest varieties (Figure 2.7
left) and were at the same time having different combinations of root growth and shoot
growth (Figure 2.7 right), that is, high root counts and high biomass (CAV4081179);
high root counts and low biomass (CAV4081278); low root counts and high biomass
(CAV4080667); and low root counts and low biomass (CAV4081444; Figure 2.7 right).
Our aim when choosing varieties by root growth in combination with shoot growth was
to maximise the chances of finding differences in the allocation of root biomass to deep
soil layers—e.g. a variety producing high shoot biomass with low root counts might be
allocating more of its root growth to deep soil layers. However, given the standard
errors of the means for rooting depth and root counts in Figure 2.6 and Figure 2.7,
genotypic differences may not be statistically significant and therefore the alleged
differences in root depth and root counts might just be due to environmental factors plus
random variation. As seen later in Chapter 4, the ranking of rooting depths and root
counts and biomass combinations do not hold from experiment 1 to experiment 3. Four
MAGIC varieties were selected but because CAV4081179 was severely lodged at
anthesis it was excluded from the analysis (Table 2.5).
2.4 Germplasm sown 61
Figure 2.6. Selection of genotypes at experiment 1 (2011) to be sown in experiment 3 (2012). Rooting depth at maturity for all varieties sampled in experiment 1 (left), and above-ground biomass as a function of the sum of root counts (root tips) along the soil profile, both at maturity as well. Coloured dots are the selected genotypes. Dots are means of 4 replicates and bars are standard errors of those means. Grey dots are the rest of the varieties sown in experiment 1, which were not selected to be sown in experiment 3.
62 Chapter 2
Figure 2.7. Selection of genotypes at 'hill plots' experiment (2011) to be sown in experiment 3 (2012). Rooting depth at maturity for the 10% deepest and the 10% shallowest roots in the 'hill plots' experiment (left), and above-ground biomass as a function of the sum of root counts (root tips) along the soil profile, both at maturity as well. Coloured dots are the selected genotypes. Dots are means of 4 replicates and bars are standard errors of those means. Grey dots are the rest of the varieties sown in the 'hill plots' experiment, which were not selected to be sown in experiment 3.
2.4 Germplasm sown 63
Table 2.5. List of genotypes, and their characteristics, sown in the three field experiments. Genotypes sown during 2011 were the same in both experiments done during that season.
Group Genotype Exp 1 & 2 Exp 3 Exp 4 & 5 Characteristics
Short (< 80 cm) TR550 Overgrowth wheat mutants: single
point mutants of the Maringa Rht1
and Rht3, BC7 isogenic lines with
varying degrees of height recovery
(see (Chandler and Harding 2013)).
Medium (> 80 < 110
cm)
MRht1
TR544
TR605
TR623
Tall (> 110 cm) MRht
TR606
TR610
Spring Espada Commercial, spring wheat, used as a
benchmark for root and shoot
growth. H45
Magenta
Pastor
Westonia
Yitpi
Long spring Beaufort Commercial, spring wheat with a
longer time to flowering in between
spring and winter wheat. Preston
Winter Revenue Winter wheat.
Wedgetail
Tin minus 5447N Near isogenic wheat pairs with
(suffixes P or plus) or without
(suffixes N or minus) the tin gene,
whose presence inhibits tillering.
6184N
6266N
6460N
Kite minus
Tin plus 5447P
6184P
6266P
6460P
Kite plus
Vigour 3-2-4 Increased early-vigour wheat: faster
leaf area expansion and, presumably,
more vigorous root growth. 7-1-1
8-5-3
W010104
W020204
W030503
64 Chapter 2
MAGIC CAV4081444 MAGIC wheat, derived from parents
Yitpi, Xiaoyan54, Westonia, SB139,
Pastor, Chara, Alsen, and Baxter. CAV4081278
CAV4080667
Triticale Bogong Triticale (× Triticosecale), which
compared to wheat has more
vigorous shoot and root growth. Currency
Contrasting
commercials
Derrimut Two commercial spring wheat
varieties that contrast in the depth of
water use: Gregory, deeper water
use, and Derrimut, less-deep water
use.
Gregory
2.5 Statistical designs
Experiments were always sown in randomized, complete-block designs, in four
replicates arranged as blocks—i.e. one replicate per block. In order to remove possible
soil or environmental effects over our response variables (e.g. rooting depth, above-
ground biomass), the spatial configuration of the experiments was further recorded by
positioning each plot within a coordinate system consisting on "columns" and "rows".
Recording rows and columns allowed spatial effects to be modelled and eventually
removed, with the resulting 'distilled' effects of the genotypes. Models used were fitted
in ASReml-R (Butler 2009)(see Appendix A).
Experiment 1 was sown in the Creek soil at the same time as experiment 2 in the Valley
soil, during 2011. They share exactly the same genotypes but of course in different
spatial configurations given by randomized allocations of the plots (Figure 2.8 and
Figure 2.9). Experiment 3 was sown in the Creek soil as well, but the number of
genotypes was reduced to 12, 4 of which were lodged at anthesis, yielding a remaining
set of 8 genotypes (Figure 2.10). Experiment 4, unlike the previous three, had an extra
2.5 Statistical designs 65
treatment that involved nitrogen fertilization of 0 or 200 kg N ha-1 (Figure 2.11). And
experiment 5 replicated the same 2 genotypes sown in experiment 4, but in combination
with two sowing densities: 100 and 200 plants m-2 (Figure 2.12).
66 Chapter 2
Figure 2.8. Statistical design of experiment 1. The experiment was sown during 2011 in the Creek site (red podzolic soil). Colour specified as 'rest' (pale green) identify plots that were not sampled, and from those, 'generic' was a wheat genotype that was used as a buffer.
2.5 Statistical designs 67
Figure 2.9. Statistical design of experiment 2. The experiment was sown during 2011 in the Valley site (yellow podzolic soil). Colour specified as 'rest' (pale green) identify plots that were not sampled, and from those, 'generic' was a wheat genotype that was used as a buffer.
68 Chapter 2
Figure 2.10. Statistical design of experiment 3. The experiment was sown during 2012 in the Creek site (red podzolic soil). Colour specified as 'lodged' (pale green) identify plots that were not sampled because they were stem-lodged at anthesis.
2.5 Statistical designs 69
Figure 2.11. Statistical design of experiment 4. The experiment was sown during 2013 in the Creek site (red podzolic soil). Two commercial wheat varieties, with supposedly different root depth at maturity, were sown in combination with 2 nitrogen fertilization treatments aimed to maximise differences in root growth: 0 and 200 kg N ha
-1 applied at the stem elongation stage
(Zadoks 31). However, and unfortunately, plots were mistakenly fertilised, and therefore blocks 2 and 4 don't have all the possible treatment combinations—i.e. in block 2, Derrimut fertilised with 200 appears in two plots, and Derrimut without fertilization none. Fortunately, nitrogen had no statistical effect over shoot or root growth, so the nitrogen treatment is later disregarded.
70 Chapter 2
Figure 2.12. Statistical design of experiment 5. The experiment was sown during 2014 in the Creek site (red podzolic soil). Four wheat genotypes where sown, but just 2 are evaluated in this thesis: Derrimut and Gregory, same genotypes as in experiment 4. Two sowing densities were sown, a low density of 100 and a high density of 200 plants m
-2. The design is a split-plot design
with 4 replicates in blocks. Experiment 6 was sown during 2014 and had the same experimental design as experiment 5, but it differed in its location: it was sown in the Quarantine site, in a yellow podzolic soil (Figure 2.1).
2.6 Field measurements 71
2.6 Field measurements
We performed several field measurements which are described in this section and
summarised in Table 2.6. Some measurements were made at discrete times (e.g. stem
elongation, anthesis, or physiological maturity), whereas others were made continuously
by data logging.
Table 2.6. List of measurements done in the field in each year. Measurements were taken at specific developmental stages (root and shoot samplings), once or twice in the whole season (bulk density, water retention parameters, water content by coring), weekly (water content by neutron moderation) or by continuous logging (water tension with gypsum blocks and canopy temperature). Stages of crop development were stem elongation (SE, growth stage 31, Zadoks et al., 1974), anthesis (A, growth stage 65) and physiological maturity (M, growth stage 87).
Exp 1 & 2 Exp 3 Exp 4 Exp 5
Measurement SE A M SE A M SE A M SE A M
Phenology
Root length density from 10 to 180 cm
depth (cm cm-3
)
Rooting depth (cm)
Shoot biomass (g m-2
)
Grain yield (g m-2
)
Leaf area index
Water tension by gypsum blocks (kPa)
Water content by neutron moderation
(cm3 cm
-3)
Water content by coring (g g-1
)
Soil bulk density (g cm-3
)
Water retention parameters
Canopy temperature (°C)
72 Chapter 2
2.6.1 Phenology
Phenological stages were recorded regularly between emergence and maturity using the
scale of Zadoks et al. (1974). Of particular interest were the stages of stem elongation
(growth stage 31), anthesis (flowering; growth stage 65) and physiological maturity
(growth stage 87). All growing stages were determined in a population basis, i.e. by
accounting for phenological events when they occurred in 50% of the individual plants
in a plot—though the border rows were excluded because they could have developed
faster given the reduced competition compared to the plants inside the plot. Growth
stage 31 was defined, according to Zadoks et al. (1974), as the start of rapid stem
growth, which was quantified visually as the appearance of the second node above the
soil surface. This cannot be seen by the naked eye—since the elongating stem is still
covered by the leaves sheets—but it can be detected by pressing with two fingers at the
base of the pseudo-stem. Growth stage 65 was determined as the date when 50% of the
spike area has protruded anthers across its spikelets. Growth stage 87 was determined
when peduncles had completed changing colour from green to yellow, when senescence
of the phloem vessels in the peduncle implies the end of grain growth.
2.6.2 Soil coring to sample roots
We sampled roots at growth stage 31 only in experiment 3, at growth stage 65 in
experiments 3 and 4, and at growth stage 87 in all experiments (Table 2.6), by taking
vertical soil cores with a hydraulic corer mounted in the three-point hitch (rear) of a
New Holland Tractor.
2.6 Field measurements 73
Soil cores were taken with tubes, 4.25 cm in diameter and 2 m in length, made of a
molybdenum-steel alloy. These tubes were pushed into the soil to a depth of 1.8 m.
Several of these cores (2 in Exp 1 and 2; 4 in the rest) were taken around the centre of
each plot. After pulling back from the soil, tubes were transported and pushed by hand
on a cradle over a table and cut at every 10 cm. The resulting 10-cm core cuts were
broken once around the middle (in Exp 1, 2 and 3) or twice (in Exp 4), once at one third
and again at two thirds of its length. Roots tips appearing in each breakage plane were
counted (i.e. two planes in Exp 1, 2, and 3, four planes in Exp 4) according to the core-
break technique described by (Drew and Saker 1980). Then the root count density, RCD
(cm-2), was determined as a ratio between the average number of root tips per plane and
the area of each plane (which was always 14.2 cm2). Roots from crops in past growing
seasons can be mixed up with the actual roots from the crop. We differentiated them by
assuming that white roots were from the current crop and brown roots from previous,
dead crops (Watt et al. 2008).
2.6.3 Soil washing to separate roots from soil
After recording the number of roots, each 10-cm soil-sample was placed into a plastic
bag, transported from the field to the lab and stored in a cold room at 5 °C. These
samples were then washed with a hydro-pneumatic soil washer built in the Australian
National University following the description made in (Smucker et al. 1982). This
machine breaks the soil cores by applying pressurized water and air and filtering the
resulting dispersed soil and floating roots (Figure 2.13). The filtered roots were later
74 Chapter 2
placed into small plastic containers with 1:1 water-ethanol solution and again inside the
cold room at 5 °C, where they were stored for no longer than 2 months.
Figure 2.13. Hydro-pneumatic soil washer used to separate roots from soil. The whole device is shown in (a). In (b), the cylinders where the soil samples are deposited are shown, this time upside-down, where one can notice the 3 pressurised-water outlets on the side of each tube's base, and the pressurised-air outlet, 1 on each tube's base. The detachable piece of each washing tube is shown in (c), and the 5-mm-mesh sieve in (d).
2.6.4 Root scanning to quantify root length density
Roots were later scanned and the length of roots per sample was determined. For that,
we dispersed the samples in water contained on an acrylic tray (27 cm × 21 cm × 3 cm)
(a) (b)
(c) (d)
Pressurised-water
outlets
Pressurised-air
outlets
Timer
2.6 Field measurements 75
adapted to fit on an A4, Epson 1680 modified-flatbed scanner. At the start of the
sampling, during 2011, we used to stain each sample by immersing the filtered roots in
a bath of Toluidine blue stain. The process was aimed at increasing the visibility of
roots under the scanner. As it was very time consuming so we sought ways of making it
faster. We tried scanning roots without staining them and compared the results of
stained and no-stain using the same samples. Root diameter and volume differed but
root length, our variable of interest, did not—or differences were so small that it was not
worth the time lost to stain them (Figure 2.14). Toluidine blue staining is recommended
for quantifying root volume and root area. However, for root length, which was our
interest, staining was unnecessary.
Later, each sample was scanned at a resolution of 400 dots-per-inch and stored as TIFF
image files. We used the software WinRhizo® (Régent Instruments Inc., Québec,
Canada) as an image analysis tool to determine the length of roots in each image. I
calculated the root length density, RLD (cm [root] cm-3 [soil]), of the sample as the ratio
between the lengths of roots measured by WinRhizo® and the volume of the sample
(which was a constant: 142 cm3).
76 Chapter 2
Figure 2.14. Comparison of stained and non-stained samples with Toluidine blue before scanning. The dashed line is y = x and the full line is the best linear fit. It can be seen that the linear fit and the dashed line fall apart in the case of average root diameter and root volume, but they are almost equal for total root length—therefore, staining the root samples with Toluidine blue does not improve the accuracy of measuring root length.
2.6.5 Predicting root length density from root count density
Because RLD is a more useful metric than RCD (for instance, one can accumulate it per
layer of soil or along the entire soil profile, whereas summing counts would not make
sense), I estimated RLD from RCD. To do that I first tried a linear model without
intercept (i.e. going through the origin, meaning that zero RCD corresponds to zero
RLD), which fitted the data well at low values of RCD but overestimated at high
values—we were counting more and more roots but the length of roots was remaining
constant. The relationship between RLD and RCD appeared to be non-linear, reaching a
plateau as RCD increased (Figure 2.15). To use the simplest model that explained best, I
fitted a rational nonlinear model with two dimensionless parameters, a, the initial slope,
and b, the curvature:
(2.1)
2.6 Field measurements 77
I found that these parameters were in turn a function of soil depth, where a in particular
decreased as the soil depth increases (Table 2.7). This means that the deeper we
sampled roots the less root length there was per root counted in the surface of the soil
sample. To estimate RLD from RCD I used the model with the corresponding
parameters according to the soil depth (Table 2.7).
For fitting these non-linear regressions, I used the "minpack.lm" R package (Elzhov et
al. 2013).
78 Chapter 2
Figure 2.15. Models used to predict root length density from root count density according to soil depth. The graph shows root length density as a function of root count density as measured in the field. Soil depths are denoted by panels, with titles in cm of soil depth. Blue lines are the non-linear fittings with equation as presented in model (2.1) and parameters and statistics presented in Table 2.7.
2.6 Field measurements 79
Table 2.7. Parameters of the equation , describing the root length
density, (cm cm-3
), as a function of the root count density, (cm-2
). Each model was fitted separately for each soil layer. Significance levels of each estimate are: 0 '***' 0.001 '**' 0.01 '*' 0.05. SE is the standard error of each estimate. RMSE is the root mean square error and R
2 is the coefficient of determination of each model.
Soil depth (cm) Estimate SE Estimate SE RMSE (cm cm-3
) R2
20 33.4 *** 8.2 5.6 ** 2.1 1.09 0.30
30 20.6 *** 2.9 3.8 *** 0.9 0.59 0.67
40 14.5 *** 2.2 2.2 ** 0.7 0.62 0.57
50 20.4 *** 4.1 4.9 ** 1.8 0.65 0.62
60 11.1 *** 1.7 1.3 * 0.6 0.71 0.68
70 16.0 *** 3.5 3.8 * 1.4 0.74 0.50
80 13.3 *** 2.6 3.4 ** 1.3 0.57 0.60
90 13.6 *** 3.1 4.9 * 1.9 0.46 0.60
100 7.4 *** 1.1 1.8 * 0.7 0.34 0.75
2.6.6 Estimation of maximum rooting depth
I determined the maximum depth reached by roots in the soil profile, RD (cm), from
data of root counts. Because maximum rooting depth is usually defined as the depth of
the deepest root that can be found, results can vary greatly due to occasional roots that
are encountered growing inside macro-pores (White and Kirkegaard 2010). Soil macro-
pores can facilitate single roots growing to great depth. To reduce this uncertainty, I
defined maximum rooting depth as the depth where crops accumulated a certain
percentage of all their roots, counting from the soil surface down the bottom of the soil
profile, RD. I chose 95% as a threshold in similarity with Trachsel et al. (2013), though
they used the threshold for cumulative root biomass rather than root counts. As roots
were sampled each 10 cm, many times the depth where crops accumulated 95% of roots
80 Chapter 2
did not match with the sampling depths (i.e. the spatial resolution). To solve this, I
interpolated within soil layers to find the depth of 95% roots. For instance, if one
particular core has 90% of its roots accumulated by 110 cm depth and 100% of its roots
at 120 cm, the depth of 95% accumulated roots would be 115 cm. I used this 95% depth
as the maximum rooting depth.
Consequently, uncorrected rooting depth was usually deeper than 95% rooting depth
(Figure 2.16). This was particularly true for rooting depths beyond 100 cm (Figure
2.16), and agrees with the notion that sometimes, by following macro-pores, individual
roots can reach depths that are unrepresentative of the whole root system (White and
Kirkegaard 2010).
Figure 2.16. The effect of considering maximum rooting depth either to be equal to the depth where the last root is found (rooting depth uncorrected) or to be equal to the depth where plants accumulate, from the soil surface, 95% of their roots. Data comes from plots sampled at anthesis (triangles), or at maturity (circles), during 2011 (pale red) or 2012 (pale blue).
2.6 Field measurements 81
2.6.7 Root length per soil depth intervals or the whole soil profile
The total length of roots along the soil profile per unit area of land, (km [roots] m-2
[soil]), was calculated as the integral of root length density ( ) along the soil profile
from the soil surface ( ) to the rooting depth ( ):
∫
(2.2)
It serves as a total account of root growth by a certain crop, but does not tell where in
the soil profile the root length was distributed. To simplify this problem but still have an
account of root length distribution along the soil profile, I also calculated values of RL
each 40 cm depth by dividing the soil profile into 3 large layers of 40 cm thickness and
a last infinite layer: 0 to 40 cm, 40 to 80 cm, 80 to 120 cm, and deeper than 120 cm.
2.6.8 Above-ground biomass and grain yield
Above-ground biomass and grain yield were determined as dry matter, at three stages:
stem elongation (growth stage 31; only in experiment 3), flowering (growth stage 65;
only experiment 3 and 4) and physiological maturity (growth stage 92; every
experiment). We collected biomass by cutting with a serrated knife all above ground
shoot, leaves and grain over an area of 0.7 m2. This area was measured by sampling
twice with a quadrat of 0.5 m width and 0.7 m long. To do so, we discarded the two
outside rows and cut biomass from the 8 inner rows of the plot. These rows were
adjacent between each other. We later dried these samples in an air-forced oven at 65
°C until constant weight was reached (around 4 days). Yield was determined by
threshing the same samples used for biomass with a stationary thresher, and then
82 Chapter 2
weighing the resulting grain as dry matter. I weighed these and the biomass samples in a
balance with an error of 0.01 g.
83
Chapter 3 Using gypsum blocks for
phenotyping water use in the field
3.1 Introduction
Gypsum blocks could make possible the continuous logging of soil water suction in
field conditions. They can measure continuously because all that is needed is a current
passing through them and a device that can record the resistance to that current, a
principle first devised by Bouyoucos and Mick (1940). Neutron probe or total domain
reflectometry methods can determine soil water content precisely, but they cannot
sample in a continuous manner. Moreover, neutron probe sampling close to the soil
surface (say, less than 10 cm depth) is risky to the operator because a substantial
fraction of the neutrons are radiated to the air and, for the same reason, measuring soil
water at these depths becomes inaccurate (Murphy and Lodge 2004). On the contrary,
gypsum blocks are advantageous not only because they offer continuous logging, but
can also be precise while cheap (Johnston 2000). However, they are not usually
employed for estimating water use by crops; rather, gypsum blocks are used for
estimating the right timing of irrigation, in perennial crops like grapes, where knowing
soil water suction is sufficient. The study of Murphy and Lodge (2004) shows how
gypsum blocks can be used successfully to predict soil water content on shallow, loose
soil layers. We are not aware of their use to estimate water use by crops deeper in the
soil, where soil is more densely structured. It is unknown, therefore, whether they can
be used for phenotyping deep water use by cereal crops.
84 Chapter 3
Previous tests of gypsum blocks properties, as well as their potential for measuring soil
suction, suggested that they were capable of measuring soil water suctions within a
range that is relevant to plants, that is, from 30 kPa (around field capacity) to 1500 kPa
(around wilting point)(Aitchison and Butler 1951). Moreover, it is known that the
readings depend on soil temperature, so a calibration to temperature was needed as well
(Johnston 2000).
The aims of this Chapter are (i) to test whether gypsum blocks could provide a reliable
measure of soil water potential under limits of interest for plants in the field, and (ii) to
see if they can be used to estimate volumetric soil water content accurately, provided
the errors accumulated after the many steps required to convert soil water suction into
volumetric water content are not excessively large.
We proceeded in six steps in order to calibrate and validate the estimation of volumetric
water content by gypsum blocks. First, we measured bulk density in two pits in the
Creek site (Figure 2.1). Knowing bulk density allowed us to convert gravimetric water
content into volumetric water content, as shown later in section 3.2. Second, to avoid
sampling every plot by direct soil coring, we measured soil water content with neutron
probe. To use the neutron probe we first calibrated it by sampling neutron counts and
gravimetric water content side-by-side under a rain-out shelter during 2012 and again,
but without a rain-out shelter, during 2013 (section 3.3). Third, by using the parameters
published by Johnston (2000), Drs John Passioura and David Deery converted the
original resistance values (Ohms) into suction values (kPa), and later did an experiment
to calculate how much that suction changes as a function of temperature (section 3.4).
Fourth, gypsum blocks were deployed in the field and measurements of soil water
3.2 Measuring bulk density 85
suction were taken almost continuously—one datum every 2 hours—during 2013 and
2014. Fifth, we converted temperature-corrected soil water suction measured by gypsum
blocks (kPa) into volumetric water content (m3 m-3). To do so, water retention curves
were determined in the lab for every soil depth where a gypsum block was buried
(section 3.6). Finally, we compared the predicted soil water content—after all the steps
required to convert gypsum block's resistance into volumetric water content—with the
soil water content measured by neutron probe as a way of validating the predictions of
gypsum blocks. We show that a surrogate soil water retention curve determined in the
field could be more appropriate than a pressure-plate water retention curve determined
in the lab (section 3.7). However, intra-experiment variability in soil physical properties
might make both approaches equally uncertain, leaving us with the only hope of using
gypsum blocks as sensors of soil water suction, not soil water content. In spite of these
uncertainties, we will venture in Chapter 5 to predict water use in two wheat varieties
that seemingly vary in rooting depth.
3.2 Measuring bulk density
Bulk density, the dry weight of soil per unit volume (BD, g cm-3), was needed in order
to convert the gravimetric water content (GWC, g g-1) into the more useful volumetric
water content (VWC, cm3 cm-3) by the following:
(3.1)
where WD is the density of water, which is approximately equal to 1 g cm-3. To sample
bulk density, we used a hand-operated device for the top 60 cm (Figure 3.1a), and a 6.5-
86 Chapter 3
cm-wide core, pushed by the hydraulic press, for depths beyond 60 cm until 160 cm
depths (Figure 3.1d). Using the hand-operated corer allowed us to sample soil more
carefully on the top, looser soil, where it is more prone to breakages given its fragile
structure. Using the hydraulic press and the core allowed us to reach depths that would
be difficult to get to by a hand corer. At depth, the soil was less fragile and retained its
integrity because soil sampled at depth greater than 60 cm had a massive, clayish
structure.
3.2 Measuring bulk density 87
Figure 3.1. Procedures followed to sample bulk density. For the top 60 cm layers, a hand corer was used, which allowed more cautioned sampling in those layers that are more prone to crumbling (a). We dug with a shovel between sampling layers 20, 40 and 60 cm depth, avoiding compaction of the soil to be sampled (b, c). For soil layers deeper than 60 cm, we cored with the same hydraulic corer as described for sampling roots, where the coring tube was actually wider (6.5 cm diameter) (d). After removal from the soil, soil samples were levelled with a scraper in order to have a precise, known volume (e).
(a) (b)
(c) (d)
(e)
88 Chapter 3
3.3 Estimating soil water content with neutron probe
Neutron probes work by radiating neutrons into the soil and counting how many are
reflected back to the source. The number of neutrons that come back varies with the
degree of saturation of the soil. It is possible, therefore, to know the soil water content
precisely from a relationship between neutron counts and soil water. There is not a
universal relationship as it varies between soils with different textures and structures.
Therefore, this technique requires a calibration to work. The calibration consists of
generating equations to directly relate volumetric soil water content (m3 m-3) to the
neutrons counted by the neutron probe. To do so in a robust way, one needs to compare
soil water content and neutron counts over a wide range of soil moistures.
Unfortunately, we failed to generate enough data at the dry end of the curve to have
precise predictions (Figure 3.2). Uncertainty in these calibrations translated to
uncertainty in the estimations of volumetric water content.
3.3 Estimating soil water content with neutron probe 89
Figure 3.2. Soil water content as a function of the neutron count ratio during 2012 and 2013 for depths of 20 cm and deeper than 20 cm. The full line shows the regression line, best fit, for both years. The dashed lines are the equations by Dr Warren Bond, detailed later in Figure 3.3.
To account for this issue, we compared our fitted calibrations to generic calibrations
developed by Dr Warren Bond, from CSIRO Land and Water, who has much expertise
working with neutron probes. Warren provided us with a generic calibration that he
produced for many different soils (Figure 3.3). According to Warren, using these
generic equations to predict soil water content is much more accurate than using an
incomplete set of data from the site of the experiments. Our calibrations were similar to
those of Warren (Figure 3.2). Given that Warren validated these calibrations in several
different soils (Figure 3.4), we used Warren's calibrations rather than ours.
90 Chapter 3
Figure 3.3. Neutron probe generic equations estimated by Dr Warren Bond. The figure shows soil water content as a function of the neutron count ratio, the ratio between the neutrons measured in the soil and the neutrons that can be counted into a tank full of water. The latter is used as a reference measurement to normalise neutron readings that could otherwise vary between devices.
3.3 Estimating soil water content with neutron probe 91
Figure 3.4. Validation of Dr Warren Bond's predictions of soil water content from neutron probe data. The graph shows measured soil water content (m
3 m
-3) as a function of neutron count ratio
(unitless) for soil beneath 20 cm at 3 locations in New South Wales: Corowa, Ginninderra and Wagga Wagga. Notice that the parameters of the curve (slope and intercept) are closely similar to those of soil deeper than 20 cm in Figure 3.3
Neutron probe samples were taken by inserting the neutron probe source into an access
tube installed in 8 plots in experiment 3 and 16 plots (all) in experiment 4. Access tubes
were made of aluminium, 2 m long and wide enough as to allow the source to be
inserted (~5 cm). Aluminium tubes were installed again by coring with the tractor,
following the same procedures as for root coring. When coring this way, the soil
removed produced a hole just wide enough to allow introducing the tube into the soil,
and although full soil contact is necessary to prevent preferential infiltration along the
external walls of the tube, the collapse of the soil around the tube that usually happens
afterwards seals the gaps.
92 Chapter 3
A neutron count ratio was calculated as a ratio between the number of neutrons counted
in the soil in each sampling and the reference neutron count that was achieved by
measuring inside a 200-L tank full of water with an aluminium access tube inside it.
This is used to normalise the neutron probe readings that usually vary from one device
to another.
Soil coring to measure soil water content directly was done at the start and the end of
the season, which, as mentioned, aimed to generate a set of equations equivalent to
those produced by Dr Warren Bond. Cores taken at the start of the season when
installing the aluminium tubes provided soil samples that were kept inside plastic bags
in 10-cm sections. These samples were transported to the lab where we weighed them
fresh (to know initial weight), and we weighed them again after drying for 48 h in an
oven at 105 °C (to know the final weight). The difference between final and initial
water content was the amount of water in the soil, which divided by the dry weight of
the sample (the final weight) gave the gravimetric soil water content. We converted the
gravimetric water content (g water g-1 soil) into volumetric (m3 water m-3 soil) by
multiplying by soil bulk density, measured in experiment 4.
As mentioned, we failed to obtain enough data points in the dry end of the curves (only
successful from 20 to 40 cm depth). However, these data were used to validate the soil
water content predictions by Warren's equations (Figure 3.5). Data and predictions
seemed to agree closely, except for soil layers 50 and 60 where the predicted data were
~5 points higher than the measured data. This could be associated with spatial variation
or uncertainty in soil bulk density.
3.3 Estimating soil water content with neutron probe 93
The agreement between predicted and measured soil water content depended on soil
depth. For 20, 30 and 40 cm depth, measured and estimated volumetric water content
agreed well; for 50 and 60 cm the estimate was greater than the measured content; and
for soil depths deeper than 60 cm, both measurements agreed on average (centre of
cloud of data points) but it has to be noted that the predicted soil water content changed
less than the measured water content, that is, for every unit change in the measured soil
water content the predicted changed less than unity. This can happen because of several
reasons. Firstly, measured soil water content includes the uncertainty of measuring bulk
density (two pits for whole trial), whereas the estimated soil water content bypasses
bulk density, for neutrons relate directly to volumetric water content (Warren's data are
fitted against volumetric water content and therefore overrides the effect of bulk
density). Secondly, installation of the aluminium access tubes may have included air
gaps between the outside wall of the aluminium tube and the soil (Evett et al. 2008).
During the installation of the aluminium tube down the hole, no slurry was added.
Instead we trusted that rainfall would collapse soil into any gaps between the aluminium
tube and the soil. This might not have happened and if air gaps were present, neutron
count would be less responsive to changes in soil water content. However, while the
sensitivity of the response function was underestimated, the average readings agreed
(centre of data clouds were the same).
94 Chapter 3
Figure 3.5. Soil water content estimated from Dr Warren Bond's equations (Figure 3.3) against soil water content calculated from gravimetric water contents at the start and end of the season of experiments 3 and 4 and the soil bulk density measured at the end of experiment 4. These comparisons serve as validations of neutron probe predictions against direct coring data. R
2 are
indicated for the linear regressions between both variables for each soil layer.
3.4 Converting resistance into suction and correcting for temperature 95
3.4 Converting resistance into suction and correcting for
temperature
Gypsum blocks are electrical resistance sensors. They consist of a porous, cylindrical
body (2.3 cm diameter by 5 cm length) made of CaSO4 within which two electrodes are
embedded (Figure 3.11d). The sensor wet and dries along with the soil, decreasing or
increasing, respectively, the resistance for current flow between the electrodes (Evett et
al. 2008). Knowing the resistance therefore solves the water suction. Two types of
commercial gypsum block sensors are available in Australia: GBLite and GBHeavy
gypsum blocks1. GBLite are appropriate to measure soil water suction in light-textured
soils, for suctions ranging between 0 and 200 kPa. GBHeavy sensors are suited for soils
with heavy texture, and measure suctions within the range 50-500 kPa. We used
GBHeavy gypsum blocks, and as seen later in the Chapter, this will impair our
appreciation of suction readings around field capacity. GBheavy sensors were calibrated
to predict soil water suction from resistance by Drs John Passioura and David Deery by
using model parameters published in Johnston (2000)(Figure 3.6).
Because electrical resistance in the gypsum blocks is also influenced by temperature, we
measured soil temperature in the field, with thermistors logging at the same time steps
and the same soil depths as the gypsum blocks (section 3.4). Resistances of a set of
gypsum blocks were calibrated in the lab against temperature by Drs John Passioura and
David Deery with pressure plate apparatus. The pressure plate and its chamber were
immersed in a temperature-controlled water bath set to 5, 15, and 25 °C. A thermistor
1 Manufactured by Measurements Engineering Australia (MEA):
http://mea.com.au/upload/Brochures/Soil%20Moisture%20Monitoring/Gypsum_Blocks_Web.pdf
96 Chapter 3
placed inside the chamber measured the actual temperature, which was slightly higher
than the temperature of the bath, perhaps because the chamber was only partially
immersed in the water bath and air was warmer (Figure 3.7 and Figure 3.8). The
increase was constantly ~1 °C for all temperature sets so it would affect the results of
the analysis only negligibly.
A set of 4 gypsum blocks was placed over the pressure plate and surrounded by a small
amount of soil to ensure good hydraulic continuity between gypsum blocks and plate.
The aim was to compare, within a range of temperatures, gypsum block readings against
the pneumatic pressure inside the chamber. At equilibrium, this should equal the actual
matric potential of the gypsum blocks. We identified equilibrium when further output of
water from the chamber was negligible and the trajectories of the gypsum block
readings reached a plateau.
3.4 Converting resistance into suction and correcting for temperature 97
Figure 3.6. Gypsum block suction (kPa) as a function of the logarithm (base 10) of its electrical
resistance (Ω, Ohms). The line shows the equation , where is the logarithm base 10 of the resistance in Ohms (Johnston 2000). The inset shows the same data in a regular axis for the electrical resistance.
Gypsum block readings are shown over time at 5, 15 and 25 °C, and at pneumatic
pressures inside the chamber of 50, 100, 200, 400, 800 (Figure 3.7) and 1500 kPa
(Figure 3.8). Readings increased dramatically when bath temperature was reduced from
25 to 15 °C, and again from 15 to 5 °C. Readings of gypsum blocks at equilibrium, the
bath temperature, and the pneumatic pressure were recorded and used, as shown in
Figure 3.9, to predict the suction reading of the gypsum block from the suction in the
pressure plate (the actual suction of the gypsum block) for each temperature. As the y-
intercepts were never statistically different from zero, the models were reduced to
slopes with a linear dependence on temperature. The equation predicting the actual
gypsum block suction from apparent gypsum block suction and soil temperature was
98 Chapter 3
suction suction
Soil temperature (3.2)
3.4 Converting resistance into suction and correcting for temperature 99
Figure 3.7. Trajectories of water suction readings (up) and the pressure-chamber temperature (down) during the calibrations. Arrows indicate the time when the readings reached a plateau (equilibrium), and hence were recorded. Numbers above those arrows (kPa) indicate the pneumatic pressure inside the chamber by that time. The colour of lines indicates each of the 4 gypsum blocks placed inside the chamber. These sets of 4 suction readings per temperature and per pneumatic pressure recorded at equilibrium are later plotted in Figure 3.9.
Figure 3.8. Idem Figure 3.7 but for a water suction of 1500 kPa. The figure was separated from Figure 3.7 to show the longer time needed for equilibrium at a greater suction.
100 Chapter 3
Figure 3.9. Suction given by 4 gypsum blocks as a function of the suction in the pressure plate at three bath temperatures of 5, 15 and 25 °C. The dashed line represents y = x. The graph on the right summarises how the slopes of the lines from the graph on the left change as a function of temperature. The latter is the equation that I used to correct suctions by temperature: y = -0.023 x + 1.472.
After correcting for soil temperature, soil water suction measured by gypsum blocks in
the field only changed slightly, as expected from equation (3.2), which states that for
every change in 1 °C in soil temperature the actual and the current gypsum block
reading will differ by ~2%. For a daily change of ~1.5 °C at 600 kPa, as shown in
Figure 3.10, a change of 18 kPa is expected, and at 1200 kPa a change of 36 kPa
(double).
3.5 Field measurement of soil water suction with gypsum blocks 101
Figure 3.10. Soil water suction, uncorrected or corrected by soil temperature (upper panel), and soil temperature (lower panel) across time for soil at 20 cm depth in experiment 1 during 2011.
3.5 Field measurement of soil water suction with gypsum
blocks
In the field, sensors were buried at eight depths in experiment 4 (20, 40, 60, 80, 100,
120, 140, and 160 cm depth) or at six depths in experiment 5 (20, 40, 60, 100, 120, and
140 cm depth). Sets of 6 or 8 gypsum blocks plus a data logger were installed in every
plot. Loggers were connected by cables to each of the 6 or 8 gypsum blocks,
transmitting the data to a central data station that later uploaded to a website. Gypsum
blocks were soaked in water overnight prior to installation. Each sensor was placed into
one hole to minimize the preferential flow of water that would be given by only one
hole for all gypsum blocks in a plot (Evett et al. 2008). Holes in the soil were made with
the same hydraulic device used for soil coring for roots but, instead, a thinner coring
tube was used, whose diameter slightly exceeded the diameter of the gypsum blocks
(~2.5 cm; Figure 3.11a and b). Soil cores were removed by pushing the tube into the
ground until the desired soil depth. Before pushing them into the holes, gypsum blocks
102 Chapter 3
were covered with a thin layer of mud (made from the same soil at the same soil depths
as the soil core just removed; Figure 3.11c). Then, gypsum blocks were pushed into the
hole with the aid of a thin wooded-stick. After reaching the bottom of the hole, around 5
cm of soil were introduced and pushed with the stick towards the gypsum block to
provide intimate contact between soil and gypsum block. Bentonite, a type of clay that
is sold commercially, was added to fill up the hole around the cable until reaching the
soil surface (Figure 3.11f). Bentonite was chosen because once wetted it expands up to
17 times, minimizing the risk of rainfall running preferentially through the walls of the
hole (Evett et al. 2008).
3.5 Field measurement of soil water suction with gypsum blocks 103
Figure 3.11. Installation of gypsum blocks and thermistors. The same hydraulic press mounted to a tractor that was used for soil coring when sampling roots was used for digging holes in the centre of plots for installing gypsum blocks, albeit a thinner metallic core was used (a, b). Gypsum blocks were soaked in water the night before (d), and covered in mud made from the same soil layer where they were supposed to be buried (c). Thermistors were deployed as well so that soil temperature could be recorded. Frame (e) shows the white cable that worked as an extension necessary to reach the desired soil depth, joint to a thermistor (black wire) with a lead sinker at the tip. After reaching the desired soil depth, the hole where either a gypsum block or a thermistor was installed was filled with bentonite to prevent the preferential flow of water along the walls of the hole (f).
Because of our need to determine soil temperature to allow for corrections of soil water
suction, thermistors, which are resistors adapted to give readings of temperature, were
installed as well. They were made of a 40 cm black-cable with the thermistor at the tip
where we added a fishing sinker made of lead (Figure 3.11e) to ensure that, by its heavy
weight, the thermistor reached the bottom of the holes. Holes were made with the same
coring tube and same procedures as for gypsum blocks. As shown in Figure 3.11e, the
(a) (b) (c)
(d) (e) (f)
104 Chapter 3
black cable (the thermistor per se) was connected to a longer white-cable with a silicone
sealed joint held together by a heat-shrinkable tube; this was done to connect the
thermistor to the data logger and also to prevent water provoking short-circuits. As with
gypsum blocks, sets of 8 thermistors were connected to a data logger in 4 plots of
experiment 4 (Figure 2.5). Soil temperature did not vary between plots.
3.6 Estimating soil water retention curves for the site
Soil water retention, also called soil water characteristic, describes how soil water
content changes as a function of the energy status of water, that is, its suction. As
suction was the reading given by the gypsum blocks, we needed soil water retention
functions for each soil and soil layer in our study in order to know the soil water
content. Knowing soil water content is important in this thesis because root water
uptake can be derived from the changes in soil water content over time.
Example soil water retentions are given in Figure 3.12. Plants can only use water held in
the soil between 10 and 1500 kPa. Consequently, as shown in the graph, the availability
of soil water to plants varies considerably from soil to soil. These upper and lower
limits for soil water holding capacity are called, respectively, field capacity and
permanent wilting point. Clay soils hold more water at high suctions but their capacity
to hold water between field capacity and wilting point is less than, say, a loam soil
(Figure 3.12). The amount of water that a sand can hold which is useful to plants is
small (Figure 3.12).
3.6 Estimating soil water retention curves for the site 105
Figure 3.12. Example water retention curves for typical soils. Data were generated from default water retention curves present in the software RETC (van Genuchten et al. 1991).
In the experiments in this thesis, changes in soil water retention functions happened
mainly between soil layers. Soil water retention can change substantially between soil
layers because of textural changes. To account for differences between layers we
sampled soil in a similar manner as for bulk density and kept samples for measuring
water holding capacity in the lab with a pressure plate. To do this, samples were taken
from two positions within the area of experiment 4: one on the north extreme and the
other on the south extreme—separated from each other by ~30 m.
To estimate the water retention functions for each soil in the north and the south
positions and in several soil layers of 20, 40, 60, 80, 100, and 120 cm, we measured soil
water content at six water suctions: 10, 30, 100, 300, 700 and 1500 kPa. For soil
106 Chapter 3
suctions between 10 and 100 kPa, soil samples were handled with caution to avoid
compaction, since soil structure is especially important for soil suctions close to field
capacity, when the most relevant pores holding water are large enough as to be affected
by compaction. For soil suctions higher than 100 kPa, when soil pores cannot be altered
by compaction, we ground the samples and sieved through a mesh that allowed us to
select only particles smaller than 1 mm; this was done to ensure appropriate hydraulic
continuity between the pressure plate and the samples, given that at these high suctions
moderate gaps between both can interrupt the water flow and thereby give a false
impression of equilibrium.
Soil samples were equilibrated in the pressure plate and chamber using two ceramic
plates according to the suction aimed: a coarser one for suctions less than 300 kPa, and
a finer one for suctions above it. Two table spoons of soil sample were placed over the
pressure plate inside aluminium rings of 1 cm depth and 4 cm wide, in sets of 16 each
time. We allocated those 16 rings as 2 soil positions (north and south), 3 soil depths
(say, 20, 40 and 60 cm) and 2 replicates per each combination of the first two;
replication was done to check for possible problems in samples reaching equilibrium.
After that, we added water slowly with a beaker so that samples wetted from the bottom
towards the top—this was done, again, to ensure hydraulic continuity between ceramic
plate and soil samples. When the top of the soil samples shined, the saturation of the
sample was finished and we applied slowly a pneumatic pressure by opening slowly the
gauge of an air-gas bottle.
When the soil samples are in equilibrium with the pressure plate at any given pneumatic
pressure, the suction in the sample is the same as in the pressure plate, which in turn
3.6 Estimating soil water retention curves for the site 107
equals the pneumatic pressure applied. As water flows out the pressure plate slowly,
equilibrium takes some time to be achieved. This time is longer the greater the suction,
as the soil water conductivity—that is, how fast the water moves along a gradient of
suction (gradient being the ratio between a difference in suction and the distance over
which that difference takes place)—decreases exponentially as the suction increases.
The pressure plate had a small outlet from which we collected the flow of water coming
out and registered the amount every hour (at low suctions at the start of the calibrations)
or once a day (at high suctions at the end of the calibration). We considered that the
samples reached equilibrium when the trajectory of the outflow coming from the
samples over the pressure plate reached a plateau. What we needed to know next was
the water content of these samples. Thus, when the plateau was attained, we quickly
released the pressure, opened the chamber, collected the soil samples and placed them
into small aluminium tins, and weighed in a scale with an accuracy of 0.01 g. Then we
placed those tins with the soil inside an oven at 105 °C for 48 h. As when measuring
soil water content by direct coring from the field (section 3.2), soil water content was
calculated as the difference between the initial wet soil and the final dry soil over the
weight of the dry soil. That gravimetric water content (g g-1) was multiplied by the bulk
density of the sample (g cm-3) and divided by the density of pure water at 20 °C (1 g cm-
3) to obtain the volumetric water content (m3 m-3) (3.1).
Once we gathered all data points of soil water content at soil water suctions of 10,
30,100, 300, 700 and 1500 kPa, we estimated the water retention curves of the form of
van Genuchten's equation (van Genuchten 1980),
108 Chapter 3
(3.3)
| |
where is the volumetric soil water content (m3 m-3), suction is given in metres rather
than kPa (1 m = 10 kPa), is the water content at saturation (m3 m-3), is called the
residual water content (m3 m-3), which is a water content small enough to be beyond the
wilting point and achieved when the function plateaus towards high values of suction,
and (m-1) and (unitless) are other curve parameters. We fitted these equations by
inputting the suction and water content data from the pressure plate into the software
RETC (van Genuchten et al. 1991). To help the fitting arrive to a solution at suctions
smaller than 10 kPa at depths of 20 and 40 cm, and smaller than 100 kPa at depths 60,
80, 100 and 120 cm, we fixed the parameter according to Paydar and Cresswell
(1996). These authors found that, for several Australian soils, a good prediction of
could be obtained by applying the formula
(3.4)
where is the bulk density (g cm-3) and is the average particle density of soils
(approximately a constant 2.65 g cm-3); the equation can be rephrased as the saturated
water content of a soil is 93% of its porosity, where porosity equals the relative volume
of soil that is not a particle (
).
3.7 Validation against neutron-probe soil water content 109
Figure 3.13. Water retention curves estimated for the site in experiment 4. The vertical axis is soil water content and the horizontal axis the soil water suction. Samples in the paddock were taken at two extremes of the experimental area: north and south. Soil water contents at suctions greater than zero were measured in a pressure plate apparatus. Soil water content at complete
saturation (i.e. suction equal zero) was estimated as
as Paydar and Cresswell
(1996).
3.7 Validation against neutron-probe soil water content
As we measured soil water content with a neutron probe within a close proximity to the
gypsum blocks (~ 1 m apart, see Figure 2.5 for experiment 4), we were able to compare
the predicted soil water content—calculated from the previous water retention curves—
with the soil water contents given by the neutron probe (Figure 3.14). Water retention
curves did not predict the water content measured by the neutron probe accurately. This
110 Chapter 3
is evident in two ways. First, the most noticeable feature of Figure 3.14 is that, at soil
depths of 20 or 40 cm, neutron-probe water contents at suctions less than 50 kPa are too
high compared to the predictions of the water retention curve fitted in the lab—reasons
for this discrepancy will be discussed later, but for now it can be said that GBHeavy
gypsum blocks are not suited for measuring suctions less than 50 kPa. Second, for
depths 60 and 100 cm, the predictions of the water retention curves were higher than the
measured water contents for the whole range of suctions. This second feature can be due
to the fact that soil sampled to determine the water retention curves in the lab might not
have been representative of the soil within the experimental area, given the high soil
heterogeneity in the field.
3.7 Validation against neutron-probe soil water content 111
Figure 3.14. Soil water content predicted by the water retention curves (red lines, same curves as in Figure 3.13), and soil water content measured by neutron probe (blue dots), as a function of the soil water suction measured by gypsum blocks.
When having a closer look at the graphs of soil water suction by gypsum blocks over
time—shown before when referring to the calibration of gypsum blocks by
temperature—one can see that even when supposedly saturated all data points are higher
than ~ 50 kPa (Figure 3.15). The principal explanation for this mismatch is that this type
of gypsum blocks (GBHeavy) were simply not manufactured for measuring soil water
suction under 50 kPa. Another type of gypsum blocks, the GBLite, would be suitable
for this task, for their range of measurement is between 0 and 200 kPa. As we were
interested in this thesis in estimating water uptake by roots, we chose to use GBHeavy
gypsum blocks, which range from 50 to 500 kPa and as such can cover more of the
112 Chapter 3
range where plants take up water. Other explanations for the lack of accuracy below 50
kPa are that some soils, depending on their pore distribution, do not necessarily reach
suction = 0 kPa even when saturated, for some soils have such small pores that even
when fully saturated water is still under tension. When the material lacks large pores,
whose negligible capillary forces when saturated would dominate the soil suction of the
whole, then the material is still under some small sort of tension even when saturated.
This may be the case for gypsum blocks. Another explanation, though less likely, could
be that sometimes air can be trapped inside pores so that the soil—or in this case, the
gypsum block—is never fully saturated. This limitation of the gypsum to read suctions
under 50 kPa is also supported by Figure 3.16, which shows the very start of the
calibration of gypsum blocks against temperature. During the 'glistering point' the soil
surrounding the gypsum blocks over the pressure plate was glistering because it was
saturated with water. Even when fully saturated gypsum blocks did not show suctions
between 0 and ~ 40 kPa.
3.7 Validation against neutron-probe soil water content 113
Figure 3.15. A closer look at Figure 3.14, narrowing the focus within soil suctions from 20 to 60 kPa. The graph shows the disagreement between the water retention curves measured in the pressure plate (red line) and the data measured with neutron probe at suctions less than 50 kPa (blue dots), where points cluster at higher values than what the pressure-plate water retention predicts (20 and 40 cm depth). The blue line shows a surrogate water retention curve that results from fitting the gypsum-neutron probe data with RETC (van Genuchten et al. 1991). The latter is however unrealistic and will be later discarded, as explained in the main text.
114 Chapter 3
Figure 3.16. The start of the calibration of gypsum blocks, where 'glistening point' shows the approximate period when the soil surrounding the gypsum blocks was shining because of being saturated with water. The graph suggests that, even when fully saturated, gypsum blocks are blind to suctions under c. 50 kPa.
Whichever the cause of the difference at suctions less than 50 kPa between gypsum-
blocks-predicted soil water content and neutron-probe-measured soil water content,
there was still a disagreement between these two quantities at high suctions (soil depths
20, 40, 60) or along the whole range of suctions (80, 100 cm). We attempted to
overcome this disagreement by fitting the van Genuchten model in RETC directly to the
neutron probe vs. gypsum blocks data, building in this way a set of surrogate water
retention curves tailored to the field data. According to what was discussed, however,
two possible surrogate water retention curves could be fitted by considering, or not
considering, gypsum blocks blind to suctions less than 50 kPa. These curves are shown
in Figure 3.17. Removing data whose suction is less than 50 kPa, albeit still using a
fixed determined as in Paydar and Cresswell (1996), affected the curves at depths 20
3.7 Validation against neutron-probe soil water content 115
and 40 cm. At a depth of 20 cm, fitting the surrogate water retention curve without data
smaller than 50 kPa produces a curve similar to the original water retention curve fitted
in the lab with the slight difference that in the surrogate one the water content at high
suctions is not overestimated (as in Figure 3.14)—which is of course redundant given
that the curve was fitted (optimized) to the data. Using all data gives a curve that passes
through the cluster of points smaller than 50 kPa, but it is so far from the water contents
predicted by the original water retention curve fitted in the lab that it seems unrealistic.
At a depth of 40 cm, the surrogate water retention curve that was fitted using all data
(i.e. even including data whose suction is less than 50 kPa) looks even more unrealistic:
for example, at a soil suction of 10 kPa this curve almost indicate that the soil is
saturated (i.e. the curve is close to the plateau) at a soil water content close to 0.40 m3
m-3 whereas both the original water retention (Figure 3.14) and the surrogate fitted
without data smaller than 50 kPa indicate soil water contents close to 0.25 m3 m-3.
Another feature of this soil water retention curve fitted with all data is that the change in
water content per unit change in suction is not smooth like the rest of the curves.
Instead, the curve describes an abrupt reduction in soil water content between 10 and c.
600 kPa, and a plateau between 600 and 1500 kPa. This lack of smoothness compared
to the other curves suggests as well that including data whose suction is smaller than 50
kPa is inappropriate. The values in the ordinates, soil water contents measured by
neutron probe, must be correct, but the values in the abscissas, soil water suction
measured by gypsum blocks, must be wrong for suctions less than 50 kPa—gypsum
blocks seem to be insensitive to smaller suctions.
116 Chapter 3
Figure 3.17. Surrogate water retention curves obtained by fitting soil water suction measured by gypsum blocks to soil water content measured by neutron probe in the field. Dots are neutron-probe-measured soil water content (circles are data < 50 kPa, triangles are suction > 50 kPa).
Notice that the parameter was fixed as in equation (3.4). Two possible surrogate water retention curves can be obtained. First, in blue, the lines show water retention curves fitted using all data. Second, in green, the lines show the water retention curves fitted after excluding data < 50 kPa. Reasons for excluding data whose suctions are less than 50 kPa are detailed in the main text and supported by Figure 3.16.
The data of neutron-probe-measured soil water content against gypsum-block-measured
soil water suction was not only clustered at around 50 kPa, but also crowded against a
ceiling upper limit of 1500 kPa. This is because the calibration of suction against
resistance shown in section 3.4 was truncated at a limit of 1500 kPa, since we did not
test the gypsum blocks beyond that suction. It is possible, though, that gypsum blocks
were reporting soil water suctions higher than 1500 kPa. That would explain why data
3.7 Validation against neutron-probe soil water content 117
points in Figure 3.14 and Figure 3.17 are clustered against a 'wall' of suction at around
1500 kPa.
Perhaps the estimation of suction by GBHeavy gypsum blocks is still valid not only at
around 500 kPa (a limit to which they were manufactured) but well beyond a suction of
1500 kPa. For instance, at 60 cm depth, if we try to guess the actual suction of the
smallest values of soil water content measured by the neutron probe at around 1500 kPa,
we could do it by extending either of the surrogate water retention curves (pretending
they are almost straight lines) until they match that value of water content somewhere
along the suction scale. That value in the scale of suction would be around 10000 kPa.
This value of suction is perhaps too high, but it is interesting that these deviations of
neutron-probe-measured soil water content from the water retention curves only happen
at soil depths where suction reaches the upper limit of 1500 kPa, that is, at 40 and 60 cm
depth; one cannot detect similar deviations at other soil depths, like 80, 100 and 120 cm,
which were never as dry as 1500 kPa. All this suggests that gypsum blocks can report
suctions much drier than 1500 kPa, but we cannot know how much or how accurately.
Possibly our water retention curves, instead of being separate curves for soil depths at
20 cm intervals, can be simplified to 4 curves: (i) at 20 cm, (ii) at 40 cm, (iii) at 60 cm,
and (iv) deeper than 60 cm. Curves at 20 and 40 cm depth are slightly different between
each other, both resembling the curves of a loam soil (Figure 3.17). The curve at 60 cm
depth matches the pattern of a clay soil, with high water contents even at high
suctions—soil water content is c. 0.25 m3 m-3 at 1500 kPa (Figure 3.17). Curves at soil
depths beyond 60 cm—that is 80, 100 and 120 cm—look slightly different between
each other when calibrated in the lab with pressure plate (Figure 3.13) but when looking
118 Chapter 3
at the surrogate water retention curves fitted to the field data they seem to belong to a
unique water retention curve. The latter agrees with our visual observation of the soil
when coring, which seemed to be homogeneously equal from 80 cm to the bottom of
the profile.
Once we had a set of putative water retention curves, both fitted in the lab or in the
field, our next step was to use each of them to predict volumetric water content from
gypsum-block's suction measured in the field. As discussed before, we discarded the
surrogate water retention curves fitted including all data given that they seemed
unrealistic. Two water retention curves were therefore considered: (i) the water
retention curve fitted in the lab, with the pressure plate data, and (ii) the surrogate water
retention curves fitted to the field data but excluding suction smaller than 50 kPa.
Then, we compared the trajectories of the calculated soil water content of both curves
plus the data points of neutron-probe-measured soil water contents. Figures for each soil
depth are shown in Appendix B. These figures show data plot by plot. Surprisingly, not
even the trajectories of soil water content predicted with the surrogate water retention
curves matched the neutron probe data very well—even against the same data that was
used for the fitting. This is indicating that the variability in the soil characteristics, and
therefore the variability in the water retention parameters, is large within the
experimental area. However, this is not unreasonable, since the water retention curves
were tailored to the whole data set, not tailored to individual plots. Soil variation can
therefore make the fittings disagree with neutron-probe-measured soil water content
from plot to plot.
3.8 Conclusions 119
3.8 Conclusions
Calibrating gypsum blocks to measure volumetric water content was a laborious process
that, unfortunately, did not yield the expected results—errors at the plot level were too
large. This can be seen in the figures in Appendix B. They show that at the plot level
neither of the two water retention curves could match the observed soil water contents
(here, I considered neutron-probe-measured soil water contents as observed values
given the proved robustness of this method). However, for the purpose of estimating
water use by plants, which is the change in water content over time rather than the
actual water content, the use of gypsum blocks might suffice. The figures in Appendix
B show that the mismatch between estimated and observed soil water contents was not a
mismatch between changes in water content. Consequently, we attempt in Chapter 5 to
calculate water use by crops by estimating changes in soil water content with gypsum
blocks.
We confirmed that gypsum blocks can report soil water suctions within the limits where
plants use water under field conditions. Our lower limit for soil water suction measured
by gypsum blocks was around 50 kPa, which is the lower limit reported by the
manufacturer and is not far from the limit of 30 kPa reported by Aitchison and Butler
(1951). Differences in the porosity of the materials used for the manufacturing of
gypsum blocks between their and our study might be responsible for this difference. In
the case of the upper limit of suction, readings from the GBHeavy gypsum blocks did
not only match with water content determined by neutron probe in the field at 500
kPa—the upper limit reported by the its manufacturer—but seemed to be valid still
around 1500 kPa.
120 Chapter 3
The fact gypsum blocks can quantify soil water status within field capacity and
permanent wilting point—though both measures are loose, not precisely defined—is
encouraging since phenotyping for deep water use could be aimed at measuring soil
water suction rather than water content. Using suction, it would not be possible to
calculate an accumulated metric over time, such as the amount of water used by crops.
But it might be possible to select crop genotypes that are successful in using deep water
by looking at some features of the dynamics of soil water suction over time. For
instance, one could quantify the onset and rate of drying of a given soil layer as
Dardanelli et al. (1997, 2004) did for the dynamics of soil water content. Under this
framework, the onset of the water withdrawal from a given soil layer might indicate
when the root length density in that layer was sufficient to make the water uptake
substantial—a sort of threshold root length density. On the other hand, once the onset of
drying started, a rate of change in soil water suction over time for that soil layer could
indicate how effective a certain genotype is in using soil water. The shape of these
putative relationships may be non-linear, since water is withdrawn at a diminished rate
over time (Passioura 1983) and on top of that the relationship between water content
and suction is exponential (Figure 3.12). Calculating the rate of change of the logarithm
of suction, rather than the rate of change if suction per se, might help to linearize these
relations and therefore have the rate of change (slope) as a parameter that summarises
how effective in using water a given genotype is. However, in order to use this
technique, sowing crops under rain-out shelters is a must, since the infiltration of rain
into the soil would disrupt the drying of the layer and make the calculation of the rate
and onset impossible.
3.8 Conclusions 121
In summary, we calibrated gypsum blocks to predict volumetric soil water content.
Neither pressure-plate water retention curves nor surrogate, neutron-probe water
retention curves were able to predict soil water accurately. However, the lower and
upper limits of available water sometimes varied in congruence—for instance, when the
lower limit was 5% higher than what the neutron probe measured, the upper limit was
5% higher as well. This indicated that predicted available water was the same. This
encouraged us to still use the gypsum blocks to measure water uptake by roots in
Chapter 5.
123
Chapter 4 Searching for genetic diversity in
rooting depth and its relationship with crop
growth and yield
4.1 Introduction
In previous Chapters I introduced the matter of deeper roots into a framework of water
use and yield (Chapter 1), I explained how we sampled roots, biomass and yield in the
field (Chapter 2), and how we developed a way of estimating soil water content
continuously in the field by using gypsum blocks (Chapter 3). This Chapter is the first
of two experimental Chapters, where I aim to address the question of whether we can
find—by sampling roots using a soil coring technique—genetic diversity in rooting
depth, and its relationship to yield, in wheat and triticale growing in the field.
The background for this Chapter is the idea that breeding grain-crops with deeper roots
might increase yield by making plants more able to capture soil resources. While this
idea is widely accepted, cases where genetic variability for rooting depth is both proved
and validated in field conditions are rare. Moreover, the presence of roots deep in the
soil does not guarantee effective use of deep water (Jordan and Miller 1980; Kirkegaard
et al. 2007); for example, roots can clump together inside macro-pores in the soil,
reducing their water uptake (White and Kirkegaard 2010).
The results of some studies suggest that deep roots have been responsible of grain yield
progress in the past. In maize, computer simulations suggest that yield increments in the
124 Chapter 4
United States corn-belt have only been possible by both more erect leaves and deeper
roots (Hammer et al. 2009). In wheat, experiments in 120-cm-deep root observation
chambers showed that genotypes developed to resist drought had longer root system in
deep soil layers (Manschadi et al. 2006). Field experiments also showed that wheat
breeding in Mexico caused crops to be more effective in extracting water from deep soil
layers (Pask and Reynolds 2013). Furthermore, breeding progress facilitated by roots
may be achieved without a yield penalty during seasons with adequate water supply, as
suggested by a computer simulation of wheat genotypes possessing a hypothetically
higher deep-water extraction capacity (Lilley and Kirkegaard 2011). These examples
support the notion that grain yield could be increased by breeding wheat varieties with
deeper root length.
Our aim in this Chapter is to test whether genetic diversity exists in current germplasm
in field conditions. To explore a wide range of phenotypes we deliberately included four
functional groups of genotypes supposed to have increased root growth. The first group
were triticales (× Triticosecale), a hybrid from wheat and rye (Secale cereale), which
grows more vigorously at early stages of crop development and in so doing could grow
more roots. Studies have demonstrated the resilience of triticale crops when compared
to wheat growing in water-limited conditions (Giunta et al. 1993).
The second group were wheat lines bred for increased early vigour. The early vigour
trait and its concomitant faster leaf area expansion has been studied in wheat by
Rebetzke and Richards (1999) and Rebetzke et al. (2004), and the root systems of these
vigorous wheats have been shown to improve early nitrogen uptake (Liao et al. 2006;
Palta and Watt 2009).
4.2 Materials and methods 125
The third group were near isogenic lines of wheat with a range of plant heights;
"overgrowth" mutants of the reduced height genes (Rht) responsible for the semi-dwarf
phenotype that made the Green Revolution possible. These genotypes could grow more
root length or depth by saving carbon that otherwise would be invested in stem mass
(Chandler et al. 2002; Chandler and Harding 2013).
The fourth group were near isogenic lines of wheat with and without the tiller inhibition
(tin) gene (Duggan et al. 2005; Mitchell et al. 2012; Kebrom and Richards 2013). The
photoassimilate saved from the production of tillers could be directed into the growth of
more roots. Considering grain yield, this advantage is double if the tillers saved were
going to be unproductive, as they usually are. A study of tin lines in tubes showed
increased partitioning to, and weight of, the root system to the five leaf stage (Duggan et
al. 2005). This increase was associated with the seminal root system; the effect on the
nodal root system, the development of which is closely associated with tiller emergence,
is unknown (Klepper et al. 1984). A study in the field showed increased root depth
associated with the tin gene, but only to the mid tillering stage (Richards et al. 2006).
We hypothesize that (i) triticales and wheat genotypes that contrast in early vigour, tiller
number and plant height, will differ in rooting depth, and that (ii) those genotypes
having deeper roots will produce more biomass and grain yield.
126 Chapter 4
4.2 Materials and methods
The majority of the details about materials and methods pertaining this Chapter were
already described in Chapter 2. But briefly, we measured rooting depth, total root
length, shoot biomass and grain yield in 3 field experiments, during 2011 and 2012 in
the Ginninderra area, Canberra, Australia (Figure 2.1).
Methods specific to this Chapter are those of statistical analyses, which were as follows.
Data were analysed by analysis of variance and linear, or non-linear, regressions with
the R statistical software (R Core Team 2014). For analysis of variance, we used linear
mixed-effects models because they allowed us to test, and account for, the effects
imposed by spatial variability of soil conditions—which, as shown later in the results
section, have a large effect over root variables. These models were fitted with the R
package ASReml-R (Butler 2009). Three different, but similar, statistical models were
used for the analysis of variance of root and shoot variables. First, there was a combined
model for experiments 1 and 2, because these two experiments share the same 36 wheat
and triticale varieties. The model was
, (4.1)
where is the response variable (e.g. rooting depth) for replicate (block)
( ), site ( = Valley or Creek), genotype ( ), column , and row
; is the overall mean; is the effect of block ; is the effect of site , is the
effect of genotype , is the effect of the interaction between site and genotype ,
is the linear effect over a gradient of columns ; is the linear effect over a gradient
4.2 Materials and methods 127
of rows ; is the column where the corresponding plot is located, and is the row
where the corresponding plot is located; is the error. Terms modelled as fixed
effects are site, genotype, their interaction, the linear trends of the column and of the
row. Random effects terms in the model are the block ( )—that is, is a
variable that distributes normally with mean equal zero and variance equal —and the
error ( )). The linear trend effects of column and row were tested for
significance and also for the influence that they have on the residuals of the model. If
without the linear effects of the column ( ) or the row ( ) the residuals
showed a trend across the column or row of the trial, but after including them (i.e.
and ) in the model the trend in the residuals disappeared, then they were
included as terms in the model. Otherwise, they were removed. When they were
included as part of the model, they accounted for the spatial trends inside the
experimental area, and, as such, helped to remove random variation from the treatment
variance: the environmental (site) and genotypic variances.
Second, we fitted a model for experiment 3, which is similar to the previous one but
lacks the effect of the site, given that the varieties in this experiment were sown in only
one site. In this regard, some varieties were shared by all 3 experiments but they were
only five, too few to draw conclusions about genotypes and genotypic groups. As a
result, we didn't analyse those five varieties across the 3 experiments. We analysed,
instead, experiments 1 and 2 combined, and experiment 3 apart. That made analysis of
experiment 1 and 2 more robust, because of the large number of genotypes (36) and 2
sites. The model for analysis of variances in experiment 3 was therefore
128 Chapter 4
, (4.2)
where the components can be understood from the previous model and an explanation
would be redundant.
Third, for analysing differences between groups of genotypes in experiments 1 and 2,
we added a fixed group term to model (4.1), and modified the genotypic term to be
random—a random subset of genotypes from within each group of genotypes:
, (4.3)
where is the effect of the group of genotype ( , that is, which is in turn a
function of the genotype ; see Table 2.5 for groups of genotypes sown in 2011).
is the effect of the interaction of the site with the group of genotypes (which in turn
depends on the genotype ).
Explicit formulae for the preceding models was written in the R programming language
and are detailed in Appendix A.
For linear regressions we used the "base" R package (R Core Team 2014). For non-
linear regressions we used the "minpack.lm" R package (Elzhov et al. 2013). Non-linear
regression was applied to predict root length density from root count density, as detailed
in section 2.6.5.
4.3 Results 129
4.3 Results
4.3.1 Differences in shoot and root traits between genotypes and
environments
As a first step to investigate differences in rooting depth between wheat and triticale
varieties, we looked into how crops distributed their root length along the soil profile,
i.e. the root length density distributions. Root length density distributions varied much
more between experiments than between genotypes (Figure 4.1). Part of the variation
was due to changes in soil textures and structures between the Creek and Valley sites in
experiment 1 and 2 (Figure 2.1), and another part was due to a greater rainfall during
2012 than 2011 (Figure 2.4). No statistical difference was detected when comparing
root length densities, at any given soil depth, between genotypes. The standard errors,
either due to errors in sampling or to the roots inherent distribution, were simply too
large. The site in experiment 2 markedly reduced root growth, both as root length per
soil layer as rooting depth. The top 20 cm of soil were similarly explored by all
genotypes, no matter in what site. For genotypes that were sown in all three
experiments, it can be seen that root length density was larger in experiment 1 than 2,
and larger in experiment 3 than 1. Because experiments 1 and 3 shared the same
location, changes in root length density between seasons can be attributed to other
factors than soil—e.g. rainfall (Figure 2.4).
130 Chapter 4
Figure 4.1. Root length density distributions across experiments and genotypes. Lines represent root length densities every 10 cm, each of which is a mean of 4 replicates. Experiments are designated as colours, as in the top-left first panel. Genotypes are detailed in the title of each panel. From those, Bogong and Currency are triticales; the rest are wheat varieties.
4.3.2 Genotypic differences in rooting depth, total root length, biomass
and yield
In this section, the root length density distributions presented previously are aggregated
(i) along the soil profile, or (ii) within soil depth intervals of 40 cm. When analysed
along the soil profile the variables under consideration are total root length (km m-2) and
maximum rooting depth (cm). When analysed within 40-cm intervals, the variables are
root length from 0 to 40 cm, root length from 40 to 80 cm, root length from 80 to 120
cm and root length deeper than 120 cm (km m-2). Between genotypes, rooting depth
differed in experiments 1 and 2 (p < 0.10; given the large errors associated with root
traits, we accept a statistical threshold of p < 0.10 as a difference worth considering) but
4.3 Results 131
not in experiment 3 (Table 4.1). Root length, as a total per m2, did not differ between
genotypes in experiments 1 and 2, but it did in experiment 3, where the triticale
genotype Currency had shorter root length than the rest of the genotypes (p < 0.05,
Table 4.1). Shoot biomass differed between genotypes in experiments 1 and 2 (p <
0.001) and experiment 3 (p < 0.05, Table 4.1). Grain yield also differed between
genotypes in all experiments (p < 0.001, Table 4.1). Both biomass and yield were
remarkably higher in triticales, particularly Bogong, than in wheat.
Between experiments, root and shoot growth were larger in experiment 1 than 2 (Table
4.1). For instance, rooting depth was deeper (p > 0.001), root length longer (p < 0.001),
shoot biomass (p < 0.01) and grain yield were higher (p < 0.001) in experiment 1 than 2
(Table 4.1).
By analysing experiments 1 and 2 together, we were able to disentangle how the groups
of genotypes, originally proposed to contrast in shoot growth patterns—namely,
triticales, early vigour, MAGIC, ‗overgrowth‘ and tin-gene wheat—differed from each
other. Groups shown in Figure 4.2 are those that seemed to differ between each other;
groups that did not differ, like early-vigour and MAGIC, are not shown. In that respect,
triticale produced higher rooting depth (p < 0.10), higher shoot biomass (p < 0.001), and
higher grain yield (p < 0.001) than the commercial spring wheat (Figure 4.2).
‗Overgrowth‘ wheat groups, grouped as ‗short‘, ‗intermediate‘, and ‗tall‘, differed with
each other for rooting depth, shoot biomass and grain yield (Figure 4.2). The ‗short‘
genotype produced higher rooting depth (p < 0.10), less shoot biomass (p < 0.10) but
higher grain yield (p < 0.05) than ‗tall‘ genotypes (Figure 4.2). Wheat isogenic line
pairs did not differ between each other when having (tin plus) or not having (tin minus)
132 Chapter 4
the tin gene, although there was a slight tendency of rooting depth and root
length to be more in tin minus, and for that reason they were included in the figure
(Figure 4.2).
Figure 4.2. (a) Root length, (b) rooting depth, (c) shoot biomass and (d) grain yield for each group of genotypes in experiments 1 and 2. Each bar is the predicted mean, generated by the ASReml model, of the genotype group when combining experiments 1 and 2. Error bars are standard errors for each mean. Above and in the middle of group bars to be compared, p-values indicate significance of their differences; where none is present comparisons are not statistically significant.
4.3 Results 133
Table 4.1. Predicted genotypic means of rooting depth, root length per unit area, shoot biomass and grain yield per experiment. Asterisks (*) or a single dot (.) indicate statistical significance, for genotype, experiment and the interaction between both, as follows: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 'NS' 1. Values within brackets are the least significant difference, calculated as twice the maximum standard error for pairwise comparisons given by the ASReml model.
Rooting depth (cm) Root length (km m-2) Shoot biomass (g m
-2)
Grain yield (g m
-2)
Group Genotype 1 2 3 1 2 3 1 2 3 1 2 3
Short TR550 112.0 72.3 11.50 4.89 1240 945 499 415
Medium MRht1 118.0 43.7 8.15 3.18 1440 1010 472 326
TR544 103.0 53.7 9.01 6.15 1430 1000 401 363
TR605 123.0 53.9 141.0 10.20 7.48 21.60 1380 1200 1800 513 379 594
TR623 106.0 44.0 8.92 5.69 1130 1210 397 385
Tall Mrht 116.0 37.5 9.54 2.68 1250 1190 411 354
TR606 114.0 31.8 11.00 3.76 1340 1160 403 364
TR610 120.0 39.0 10.00 6.75 1230 1240 408 354
Spring Espada 124.0 37.6 8.84 2.77 1370 995 564 369
H45 120.0 56.5 11.60 3.58 1420 1270 573 472
Magenta 118.0 31.5 10.30 4.57 1180 761 536 401
Pastor 108.0 54.0 10.20 5.15 1340 950 567 393
Westonia 125.0 48.6 11.30 3.88 1480 1110 602 405
Yitpi 117.0 39.0 9.57 2.50 1210 968 551 397
Long spring Beaufort 113.0 75.0 8.46 8.12 1360 869 583 432
Preston 104.0 72.2 9.75 4.54 1370 1170 615 497
Winter Revenue 103.0 45.3 8.65 4.17 1270 1280 650 536
Wedgetail 108.0 67.9 9.14 6.22 1330 920 629 357
T in minus 5447N 134.0 69.3 11.20 3.71 1240 713 495 423
6184N 114.0 50.4 137.0 11.00 3.68 23.70 1340 742 1940 635 300 695
6266N 118.0 70.8 12.00 5.58 1100 1080 453 427
6460N 121.0 51.4 12.10 5.29 1240 799 574 407
Kite minus 138.0 41.2 12.40 4.49 1180 960 506 352
T in plus 5447P 124.0 61.0 8.78 6.14 1140 868 553 372
6184P 124.0 69.5 134.0 11.70 4.92 21.20 1310 860 1890 571 372 666
6266P 117.0 57.1 12.90 3.79 1320 891 473 410
6460P 105.0 36.6 10.10 3.28 1080 778 512 358
Kite plus 99.5 54.8 8.32 6.01 1190 1070 459 458
Vigour 3-2-4 113.0 64.0 8.65 2.91 1120 812 452 330
7-1-1 91.4 53.8 8.65 4.54 1170 884 438 369
8-5-3 106.0 62.9 10.20 5.63 1190 852 440 318
W010104 129.0 43.1 7.26 6.35 1450 645 541 337
W020204 115.0 47.1 10.10 3.58 1220 578 464 339
W030503 104.0 52.5 6.96 3.22 1220 986 426 407
MAGIC CAV4080667 128.0 20.30 1670 484
CAV4081278 148.0 22.00 2000 739
CAV4081444 134.0 21.20 1800 604
134 Chapter 4
Triticale Bogong 130.0 82.9 145.0 11.40 5.20 20.60 1660 1360 2070 792 651 714
Currency 118.0 47.3 137.0 8.90 4.37 18.30 1360 1270 1900 595 522 628
Genotype (G) ∙ (26.5) NS NS *
(2.76) *** (216)
*
(190) *** (93)
***
(92)
Experiment (E)
*** (5.5) *** (0.74) ** (110) *** (36)
G×E NS NS *** (353) * (149)
4.3.3 Relationships between rooting depth, total root length, biomass
and yield
Shoot and root traits, gathered together across the three experiments, correlated with
each other (Figure 4.3). The greater the rooting depth, the greater the shoot biomass (p <
0.001, Figure 4.3a) and grain yield (p < 0.001, Figure 4.3c). The relationships were
curvilinear, and thus while moving from 50 to 100 cm depth corresponded with a
biomass increase of ~3 t and a yield increase of ~1 t, changing rooting depth from 100
to 150 cm depth corresponded with a change in biomass of ~8 t and a change in yield of
~2 t.
Shoot biomass and grain yield also correlated with total root length (Figure 4.3b and
Figure 4.3d). The larger the total root length, the greater the biomass (p < 0.001, Figure
4.3b) and the greater the yield (p < 0.001, Figure 4.3d). This time the relationships were
better represented by a straight line.
At the same rooting depth or root length, triticale genotypes produce more biomass and
yield than wheat genotypes (Figure 4.3). This was more apparent in the case of yield,
where triticale, at the same rooting depth or root length, produced ~1 t more yield than
wheat in experiments 1 and 2 (Figure 4.3). Differences between biomass and yield
4.3 Results 135
between triticale and wheat having the same rooting depth or root length were smaller in
experiment 3 (Figure 4.3).
Figure 4.3. Relationship between shoot and root variables at physiological maturity. (a) Shoot biomass, or (c) grain yield, as a function of rooting depth, the maximum soil depth reached by roots, and (b) shoot biomass, or (d) grain yield, as a function of root length, the total root length per unit area. Points are the predicted means, generated by the ASReml model, for each wheat (circles) or triticale (triangle) genotype in each experiment. In (a) and (c) the quadratic term was significant and thus a quadratic model was fitted. In (b) and (d) it was not significant, thereby a linear model was fitted.
When analysing all experiments together, root length deeper than 120 cm was directly
related to shoot biomass and grain yield, but not harvest index (Figure 4.4). When
fitting these relationships per experiment rather than altogether, root length deeper than
120 cm was directly related to shoot biomass (p < 0.05), grain yield (p < 0.01) and
harvest index (p < 0.01) only in experiment 3 (Figure 4). In experiments 1 and 2 these
136 Chapter 4
relationships were not statistically significant (Figure 4.4). The more root length was
produced in that deep layer the more biomass and yield crops produced in experiment 3
(Figure 4.4). Correlation does not imply causation, but shoot biomass and grain yield
were larger in experiment 3 than experiment 1 and 2, and root length in that deep soil
layer was larger as well. It is likely that crops in experiment 3 had more available water
in the subsoil and thus more root growth there produced more above-ground growth.
The relationships did not seem to be driven by genotypes in experiments 1 and 2, but
they did in experiment 3. In that experiment, shoot biomass and grain yield correlated
closely with deep root length (p < 0.05 and p < 0.01, respectively), and, given that data
are predicted genotypic means, the variability is mostly caused by genotypes. This
means that genotypes drove the relationship between deep root length and shoot and
grains growth (Figure 4.4).
4.3 Results 137
Figure 4.4. Shoot biomass (left), grain yield (centre) and harvest index (right) as a function of the root length per soil layer (rows) in each of the three experiments. Points are the predicted means, generated by the ASReml model, for each genotype in each experiment. R
2 is the
adjusted R-squared and p is the p-value for the straight-line fit relating either shoot biomass or grain yield with root length deeper than 120 cm only at experiment 3. Note the change in scale in the x-axis across depth intervals.
138 Chapter 4
4.3.4 Plant height, root and shoot growth
To further investigate the differences in rooting depth, biomass and yield between
‗short‘, ‗intermediate‘ and ‗tall‘ wheat genotypes, we related those variables to canopy
height measured just after anthesis (when the maximum height of the crop is achieved).
There were trends suggesting a direct relationship between shoot biomass and plant
height (p < 0.07) and a negative relationship between rooting depth and plant height (p
< 0.06; Figure 4.5). Grain yield correlated negatively with plant height (p < 0.02; Figure
4.5). However, the shortest genotype, TR550, represented a leverage point for the linear
fittings in Figure 4.5: without it there would not be any trends. Nevertheless, the trends
seem to suggest that wheat with shorter stems have less total above-ground growth
(biomass) but higher yield and deeper rooting depth.
4.4 Discussion 139
Figure 4.5. Predicted genotypic means, modelled from experiments 1 and 2 combined, of rooting depth (a), shoot biomass (b), and grain yield (c) as a function of final canopy height achieved by crops after anthesis. Bars are the standard errors of the predicted means. The 8 genotypes evaluated are those differing in point mutations for the dwarfing genes (Chandler and Harding 2013). Lines represent the best fit with the following statistics: (a) R
2 = 0.47, p < 0.06;
(b) R2 = 0.44, p < 0.07; (c) R
2 = 0.63, p < 0.02.
140 Chapter 4
4.4 Discussion
We found that triticale and wheat genotypes differed in rooting depth when evaluated in
field conditions. However, differences were much larger between experimental sites or
years than between genotypes. This may be due to the fact that soil physical properties,
water and nutrients availability have a great influence over root growth which could
override any genotypic effect. In order to detect differences between varieties, this study
suggests that a greater effort has to be placed in making proper statistical designs that
can account for soil and season variability.
We found, however, differences in rooting depth between genotypes. These differences
were most evident when comparing different species, that is, triticale and wheat. Indeed,
triticales produced deeper roots and above-ground growth than commercial wheats.
Although introgressing these characteristics of deep roots from triticale to wheat is
challenging, the results are encouraging since they support the notion that a more
vigorous shoot, by faster leaf area expansion and biomass accumulation, can increase
root growth. This breeding objective was already undertaken by (Rebetzke and Richards
1999; Rebetzke et al. 2004). However, the early vigour varieties that we evaluated in
this thesis did not produce deeper rooting than other wheats. This may be due to the fact
that these varieties were not particularly high biomass producers, which can diminish
root growth; they had a more dwarf type of plant (genotypes 7-1-1 and 8-5-3) and a
planophyle leaf structure (that is, leaves closer to an horizontal rather than vertical
angle) which could help in preventing direct soil evaporation by quickly covering the
soil, but were not producing high biomass and it seems that neither high root length and
depth.
4.4 Discussion 141
Between wheat genotypes, differences in rooting depth were only noticeable amongst
those genotypes that through mutations in the Rht genes showed a range of plant
heights. For breeding purposes the genetic mechanisms behind these changes are
promising because they were given by a change in only one gene. This is not only
interesting because of genetic simplicity and its consequent ease of breeding, but also
for the possible mechanism behind. It is likely that shorter genotypes were producing
deeper roots and more yield by taking advantage of the same trade-off that during the
Green Revolution increased yield in wheat, that is, by sacrificing stem growth and
favouring grain growth and in this case root length as well. If this mechanism is truth it
may still be possible to reduce excessive growth of stems in tall plants while increasing
root growth. This should also be true in the case of the prevention of extra tillers by
means of the tin gene. But unlike Duggan et al. (2005), reducing the growth of extra
tillers did not increase root growth and yield in our field experiments. These authors,
however, found differences in root growth in a wheat isogenic line pair that we did not
evaluate (Banks). There were no differences when they tested the isogenic lines Kite +
tin against Kite, a pair that we did evaluate in this study. Most important perhaps is the
distinction that Duggan et al. (2005) measured root growth in tube experiment in 50-
days old plants whereas we did it in the field in plants that reached maturity. Differences
in root growth between isogenic line pairs with and without the tin gene might
disappear towards older stages of crop development.
Crops with deeper roots certainly yielded more. But because one's aim is to improve
yields by breeding, one must wonder how much of the variation in rooting depth and
yields was provoked by genotypes. The answer is that it was not much. Most of the
142 Chapter 4
variation was caused by environments. For example, in Figure 4.3 yield and biomass
were both related to rooting depth and total root length, but data points were clustered
according to the experiment where they belonged. Within each cluster, i.e. within each
experiment, genotypes did not drive the relationships (that can be seen in the lack of
correlation between a given abscissa value and a given ordinate value per experiment).
This means that rooting depth and total root length were associated with biomass and
yield, but genotypes had little responsibility over these relationships—environmental
factors had greater impact.
Genotypic effects on the relationship between deep root length and crop above-ground
growth might have taken place in experiment 3. These can be seen in Figure 4.4, where
only in experiment 3 there were significant regressions between shoot biomass, grain
yield and harvest index with the length of roots beyond 120 cm depth in the soil. Data
points are distilled, predicted genotypic means and therefore the source of variation
comes mostly from genotypes. Breeding for more root length deep in the profile, at least
according to experiment 3, can increase biomass and yield.
4.5 Conclusions
Increasing rooting depth has been proposed as a way of increasing wheat yield, both in
water limited and in high yielding conditions. However, examples where it differs
between genotypes growing in field conditions are rare. We showed that rooting depth
and root length deeper than 120 cm can vary between triticale and wheat genotypes and
these differences in turn correlate with biomass, harvest index, and grain yield. Between
genotypes, the greatest differences in root and shoot growth were between triticale and
4.5 Conclusions 143
wheat. Within wheat, the greatest differences were between isogenic lines that differ in
point mutations conferring them a range of plant heights. In that respect, a trend is
suggested, where shorter plants produced less biomass but greater rooting depth and
yield, which could signify that a reduction in stem growth results in more assimilates
being available for root and grain growth. To corroborate the existence of this trade-off
more experiments should be conducted, where several wheat isogenic lines (more than
the number tested in this study) are tested in the field for differences in stem biomass
and root growth and root depth. Ideally, these lines should vary along a gradient of plant
heights so that several lines fall under the ‗short‘, ‘intermediate‘ and ‗tall‘ classes, for
assessing only one line (TR550) in the ‗short‘ class proved to be a weakness of the
analysis in this study. Validating the existence of this negative relationship between
stem biomass and root growth would conform with the concept of ideotypes by Donald
(1968) and encourages the exploitation of this trade-off under other circumstances, like
reducing tiller number (Mitchell et al. 2012) or overproduction of roots in shallow soil
layers (Zhang et al. 1999).
In summary, there were genotypic differences in rooting depth, they correlated with
grain yield, and we were able to detect them by soil coring. Soil coring was challenging
because it was time consuming and genotypic differences were elusive because of the
strong environmental effects over root growth. The careful design of experiments
accounting for spatial variation in soil properties to distil the genotypic effects from the
soil effects may prove highly useful in phenotyping for deep rooted genotypes in the
field.
145
Chapter 5 Relating deep-root length with water
use, canopy temperature and grain yield
5.1 Introduction
Whether we can detect genotypic differences in rooting depth in current germplasm
growing in field conditions, and if those differences have value for grain yield, was
addressed in the previous Chapter 4. That was done by sampling roots by soil coring, a
technique that is of utmost importance but is very time consuming. In this Chapter I aim
to find an alternative, or a complementary approach, to detect deep roots by measuring
canopy temperature, as well as analysing the relationship between canopy temperature
and water use determined with the gypsum blocks sensors, whose calibration was
previously detailed in Chapter 3. This approach is important if we are to incorporate the
selection of deep roots into breeding programs.
In farming systems where soil water limits grain yield, crops with deeper roots are
expected to yield more grain if water is available deep in the soil. Nevertheless,
selecting for them by directly quantifying roots in the field is too slow and it lacks the
precision required by breeding programs. Although high throughput soil coring methods
have been developed in the field that can assess rooting depth in large numbers of
genotypes or treatments, the soil spatial variability even within the same general area is
an issue which requires greater replication or smarter designs to overcome (Wasson et
al. 2014). Hence, we need to develop high-throughput field phenotyping methods for
146 Chapter 5
deep water uptake by roots if we are to incorporate this trait into crop improvement
programs. There are several possibilities that may be useful proxies for rooting depth.
The most widely discussed ones are canopy temperature (Olivares-Villegas et al. 2007;
Reynolds et al. 2007; Lopes and Reynolds 2010; Pinto and Reynolds 2015) and green
leaf area retention (Borrell et al. 2000; Christopher et al. 2008), both of which should be
related to root systems being able to access deep soil water. Differences in canopy
temperature are probably most feasible due to the availability of cheap infrared
thermometers and thermal imaging cameras that can be used. These are useful because
the temperature of a canopy results from a balance between heat gain and heat loss, and
a large fraction of that loss is latent heat associated with evaporation from the leaves
(transpiration). Hence, cooler canopies indicate higher transpiration rates and greater
stomatal conductance (Choudhury et al. 1986; Dunin et al. 1991) and thereby an
effective root system with access to soil moisture. Although water captured anywhere in
the soil profile can supply transpiration, there are examples in wheat where cooler
canopies correlated specifically to deep water use and deep roots in landraces and
genotypes belonging to a breeding population (Reynolds et al. 2007; Lopes and
Reynolds 2010; Pask and Reynolds 2013).
In water-limited environments, water used by a wheat crop is often more valuable after
anthesis, when most of the current assimilates produced by the crop are partitioned
directly to grain growth. Indeed, the efficiency with which wheat converts water use
into grain is around 22 kg of grain ha-1 mm-1 over the whole season (French and Schultz
1984; Sadras and Lawson 2013), but can be more than doubled, between 55 (Manschadi
et al. 2006) and 59 kg of grain ha-1 mm-1 (Kirkegaard et al. 2007), during the grain-
5.2 Materials and methods 147
filling period. Therefore, if the top soil is dry but water is available at depth a wheat
crop with deeper roots during grain filling is expected to have a cooler canopy, use more
water, and produce more grain through a higher-than-usual water-use efficiency.
A low canopy temperature can therefore be used as a surrogate for roots accessing soil
water at depth in some environments. Taking spot measurements of canopy temperature
in field plots is useful to provide a snapshot of variation but it does not give dynamic
information on canopy temperature over time. Moreover, canopy temperature depends
on air temperature and vapour pressure deficit. To account for the effect of both, we
used the approach of calculating a crop water-stress index as proposed by Idso et al.
(1981) and Idso (1982) and used elsewhere (Cárcova et al. 1998). In this Chapter, I
have studied the dynamics of canopy temperature, the crop water-stress index, and soil
water status of two contemporary wheat varieties grown in Australia known to vary for
their rooting depth in the field at maturity. We report this together with variation in
biomass and yield. We hypothesize that (i) continuously monitored canopy temperature,
and expressed in the form of a water-stress index, can be used to detect deep water use
in high-yielding wheat varieties, and that (ii) the wheat genotype shown to have deeper
roots will have a cooler canopy temperature and deeper water use, which will result in
more above-ground growth, including grain yield.
148 Chapter 5
5.2 Materials and methods
5.2.1 Growing conditions
The experiments were designated as experiment 4 (2013), and experiment 5 (2014), and
were conducted in the field in the Ginninderra region, ACT, Australia (Figure 2.1).
Plots were machine sown on the 5th June in both 2013 and 2014, and were 6 m long and
1.8 m wide. Each plot was 10 rows wide with 0.18 m between rows. This is typical for
wheat growing in this region. Plots were fertilized, in all experiments at sowing, with a
mix of 15-12-0-12 Croplift 15® (percentages of N, P, S and K, respectively) at a rate of
100 kg ha-1. Other abiotic (nutrients) or biotic (weeds, diseases) stresses were avoided
by sound management, since we aimed to achieve growing conditions where only
radiation, temperature and water limited growth of the crops.
Air temperature and vapour pressure deficit were measured with a weather station next
to the experiments. Vapour pressure deficit (VPD) was calculated as the difference
between the saturated vapour pressure that would correspond to the actual air
temperature and the actual vapour pressure in the air, which was registered by the
weather station. Rainfall was registered by the weather station as well.
5.2.2 Measuring canopy temperature
Canopy temperature was measured with 2 infrared thermometers positioned on one side
of every plot, pointing at their centres with a North direction (Figure 5.1). The infrared
thermometers had a built-in data logger that registered one datum every hour.
Thermometers were first calibrated against a black body apparatus that served as a
5.2 Materials and methods 149
reference temperature. Then, canopy temperature, air temperature, and VPD were
matched together according to time of measurement and were used to calculate canopy
temperature deviation (CTD), the temperature of the canopy minus the temperature of
the air (°C); the more negative the CTD the cooler the canopy with respect to the air.
This is an important distinction as other researchers have used 'canopy temperature
depression', which equals the temperature of the air minus the temperature of the
canopy (Amani et al. 1996). Using 'canopy temperature deviation', as we do, is
consistent with the pioneering work of Idso et al. (1981), and gives us the chance of
further calculating a crop water-stress index.
150 Chapter 5
Figure 5.1. Pictures showing the infrared thermometers used to measure canopy temperature hour after hour. On the left, it can be better appreciated that they were positioned in a 45° angle from the vertical line. On the right, it can be seen more about its contruction: a PVC tube inside which the electronic components are installed, and a frontal aluminium head with a hole in the middle where the infrared sensor lays.
5.2.3 Calculating a crop water-stress index
To normalise CTD by VPD, we calculated a crop water-stress index as proposed by
Idso et al. (1981) and Idso (1982). Crop water-stress index (CWSI) will adopt values
between 0 (no stress) and 1 (fully stressed), and is defined as
(5.1)
where CTD is the current canopy temperature deviation on a given day (°C); CTDLL is
the CTD lower limit (°C), representing the non-water-stressed baseline; and CTDUL is
the CTD upper limit, which is the CTD that would be achieved when transpiration is
zero (°C). Therefore, CTDLL is the minimum CTD that can be achieved by a well-
watered crop at the current level of VPD. This would ideally be measured in a well-
watered crop in our experiments, but we lacked that type of plot. As a substitution, we
5.2 Materials and methods 151
used a prediction equation determined by Idso (1982) for wheat, which, as shown later,
matched our data well (Figure 5.8). The prediction equation was
(5.2)
where is canopy temperature (°C), is air temperature (°C), and
, as detailed in Idso (1982 p. 62). While this equation was
originally developed for wheat during pre-heading, and our canopy temperature data are
for post-heading, using this pre-heading equation matched our data better than a post-
heading one. Therefore, we used this curve. Reasons for this disagreement could be due
to the fact that our more modern wheats are retaining more green leaf area during post-
anthesis than the durum wheat (Triticum durum) that Idso (1982) used for his
calibration.
CTDUL is the maximum CTD that can be achieved when transpiration is nil. While this
cannot be measured experimentally, it can be estimated from equation (5.2) as
(5.3)
where VPD' is vapour pressure deficit determined with help of parameter , as in the
following equation
(5.4)
where is vapour pressure determined at a temperature equal to , and
is vapour pressure determined at air temperature. VPD' is therefore only a function of
air temperature. The concept of CWSI is represented graphically in Figure 5.2.
152 Chapter 5
Figure 5.2. Graphical depiction of the concept of crop water-stress index (CWSI) developed by
Idso et al. (1981) and Idso (1982). The CWSI is a ratio between how far the current CTD of a crop is from the non-water-stressed CTD, and the maximum possible difference in CTD with the non-water-stressed CTD. The current CTD of the crop is exemplified in the graph as a black point. The non-water-stressed CTD (CTD lower limit = CTDLL) is determined by equation (5.2) and is represented by the blue line. The maximum CTD at the current air temperature (CTD upper limit = CTDUL) is the top grey line. The segment representing the distance between CTD and CTDLL is the numerator, and the segment between CTDUL and CTDLL the denominator, of equation (5.1).
5.2.4 Calculating water use and the fraction of plant available water
Water use was determined from the gypsum blocks data, as shown in Chapter 3, where I
described how we calibrated the gypsum blocks to read volumetric water content
(VWC) of the soil every 2 hours at several soil depths. Then, from the trajectories of
VWC over time, we calculated the mean VWC per layer for each day. Because studying
the system in one dimension is enough, we calculated the millimetres of water that
corresponded to each soil layer by multiplying VWC by the thickness of the soil layer
5.2 Materials and methods 153
(200 mm). We established that every negative change in mean VWC (or mm of water)
between two successive days equalled root water uptake and soil evaporation (i.e.
evapotranspiration = water use). This has the drawback that water use can be confused
with internal soil percolation, but as rainfall events were not large (Figure 2.4) we
assumed percolation was negligible. Because our measuring time was post-anthesis, it
can be assumed that, under full canopy cover, direct soil evaporation would be nil, and
water use would be equivalent to transpiration, that is, root water uptake.
The fraction of plant available water (FPAW) was calculated as the ratio between the
actual plant available water at a given time (PAW) and the maximum PAW that the soil
profile can hold (PAWmax)
(5.5)
Both PAW and PAWmax were initially calculated per soil layer and later for the whole
soil profile by adding the individual values of every soil layer until a depth
corresponding to the maximum rooting depth of the corresponding trial (140 cm in
2013, and 100 cm in 2014). For each individual layer, PAW was estimated as the
amount of water held above wilting point. Wilting point was estimated as the VWC
when soil water suction is 1500 kPa. Also for every soil layer, PAWmax was estimated
as the amount of water than can be hold between field capacity and wilting point, where
field capacity was estimated as the VWC when soil water suction was 30 kPa. To
calculate the VWC at each of those suctions, we used equation (3.3).
154 Chapter 5
5.2.5 Measuring root length density along the soil profile
Root length density was estimated by counting roots by the core-break technique,
performing root washing and root scanning (as detailed previously from sections 2.6.2
to 2.6.5).
5.3 Results
5.3.1 Growing conditions
Seasons differed in water availability, as 2014 was drier than 2013 (Figure 2.4). Rainfall
was less during spring in 2014 than 2013. Both seasons, however, had more rain than
the average season when compared to long term means from 1889 to 2014 (Figure 2.4).
Both seasons were similar in terms of thermal regimes (Figure 2.4).
5.3.2 Root length density distributions
Root length densities were greater in soil layers near the surface than deeper in the
profile. The maximum depth reached by roots was higher in 2013, probably in response
to a wetter spring (Figure 2.4). Variability in root length density was high, and
differences between genotypes were not statistically significant in any soil layer (as
evidenced by the large standard error bars, Figure 5.3). However, in all three
experiments, Gregory had consistently higher mean root length densities below 1 m
depth than Derrimut (Figure 5.3).
5.3 Results 155
Figure 5.3. Root length density distributed along soil depth for the varieties Derrimut and Gregory during 2013 (left) and 2014 (centre and right). The complete profile is plotted on top, and a subset, at 100 cm and deeper, is plotted at the bottom for better appreciation of small root length density values. Points are means, and horizontal bars are standard errors, of 8 replicates. For data from 2014, an extra site is shown (right). For that extra site only data of root length density is shown in this thesis, since air temperature and soil water characteristics were not measured, making estimations of canopy temperature deviation and soil water content inaccurate.
5.3.3 Canopy temperature and water use
Canopy temperature deviation from air temperature (CTD) around solar noon was
consistently more negative for Gregory than Derrimut, for both 2013 and 2014 (Figure
5.4), indicating that Gregory's canopy was cooler than Derrimut's. Before anthesis, CTD
around solar noon was closer to negative values than after anthesis. CTD became
progressively more positive towards maturity while VPD was increasing. This may
have been partly caused by senescence of leaf area, a quantity that we did not measure.
Mean CTD around solar noon was similar between both genotypes during a rainfall
156 Chapter 5
event, since rain would have cooled down the foliage to a common temperature—e.g.
most evident the 8th, 10th and 11th November 2013, and 15th, 16th and 24th November
2014. When rain was largely absent, as it was for long periods pre- and post-flowering
in both years, CTD became more negative as water use per day increased. CTD also
responded to VPD, becoming more negative as the latter increased—e.g. between 4th
and 6th October, and between 7th and 10th October 2013 (Figure 5.4).
Figure 5.4. Trajectory of canopy temperature deviation from air temperature (CTD) at solar noon ± 2 hours, during 2013 and 2014 (top); water use per day during 2013 and 2014 (middle); vapour pressure deficit (VPD) at solar noon ± 2 hours; and rainfall per day during 2013 and
5.3 Results 157
2014 (bottom). Rainfall bars are stacked on top of each other when coinciding between years. Anthesis and physiological maturity dates are indicated, respectively, by a vertical, dashed line the 22
nd October and another the 12
th December. For CTD and water use, points are the means,
and bars the standard errors, of 8 replicates.
Water use per day was greater in many instances for Gregory than Derrimut, both
during 2013 and 2014 (Figure 5.4). During 2013, water use per day had two peaks, one
at the beginning of booting (Zadoks 40; Zadoks et al. (1974)) and another at mid-grain
filling (at around Zadoks 80; Zadoks et al. (1974)), during the end of November (Figure
5.4). In both peaks water use per day surpassed 4 mm day-1. During 2014, water use per
day had one main peak just after anthesis, during the end of October and the start of
November. Values were less than during 2013, never reaching 4 mm day-1.
158 Chapter 5
Figure 5.5. Relative water use per soil layer and per day for Derrimut and Gregory during 2013 (top) and 2014 (middle), and rainfall per day (bottom). Water use is relative to the total water use per day, and it is indicated by coloured areas. Numbers inside those areas are their soil depths, in cm. Rainfall bars are stacked on top of each other when coinciding between years. Vertical, dashed lines across panels indicate anthesis (left) and physiological maturity (right) dates as in Figure 5.4. Relative water uses per soil layer are means of 8 replicates.
In addition to typically capturing more water, Gregory used relatively more water from
deeper soil layers than Derrimut (Figure 5.5). During 2013, Gregory took up
proportionally more water from the 140 and 160 cm layers than Derrimut (Figure 5.5).
These results of day after day relative water uptake were confirmed later, when water
5.3 Results 159
use per soil layer was accumulated for the whole grain filling period (Figure 5.6).
Accumulated water use in Gregory was greater than Derrimut at depths of 100 and 140
cm (p < 0.05, Figure 5.6). Moreover, these data given by gypsum block sensors were
supported by data given by neutron probe, which showed a slightly higher (but not
statistically significant) water uptake of Gregory between 100 and 140 cm depth (Figure
5.7). During 2014, root water uptake was shallower than 2013 for both varieties, but
Gregory, again, had proportionally more water uptake at the deepest, 100 cm, depth
than Derrimut (Figure 5.5). Those differences were however non-significant when
accumulated during the whole grain-filling period, where mean water use was
seemingly larger in Gregory but not statistically significant from Derrimut (Figure 5.6).
Unfortunately, we did not use a neutron probe during 2014 that could confirm our lack
of differences in accumulated water use estimated from gypsum blocks.
160 Chapter 5
Figure 5.6. Accumulated water use from anthesis to maturity along the soil profile for Derrimut and Gregory during 2013 (left) and 2014 (right). Accumulated water use was determined as the sum of every negative change in volumetric water content (m
3 m
-3) measured by gypsum blocks
between two successive days. Points are means, and horizontal bars are standard errors, of 8 replicates. Asterisks at soil depths 100 and 140 during 2013 indicate statistically significant differences (p < 0.05). Other soil depths during 2013 and 2014 did not differ statistically.
5.3 Results 161
Figure 5.7. Soil water content along the soil profile, at the start of the season (black line and triangles), and at maturity of each genotype (coloured points and lines) measured by neutron probe during 2013. Points are the means, and horizontal bars the standard errors, of 8 replicates in the case of soil water content at maturity. Initial soil water content is a mean of all plots in the trial, as they did not vary appreciably.
5.3.4 Crop water-stress index
Because CTD is affected by VPD, we followed the method of Idso et al. (1981) and
Idso (1982) for calculating a new variable that normalizes the effects of both air
temperature and VPD: the crop water-stress index (CWSI), as explained in section
5.2.3.
When our data of CTD was placed within the framework of Idso (1982), it matched
well, as the field data fitted within the lower (CTDLL) and upper limits (CTDUL) (Figure
5.8). This gave reassurance that predicting CTDLL from the model published by Idso
(1982 p. 62) was appropriate. There were, however, some values of CTD that were
more negative than the CTDLL baseline (Figure 5.8). This could be due to the effect of
162 Chapter 5
variables not included in the calculation of CWSI that can still affect CTD, like wind
speed. The upper limit of CTD, the CTDUL, increased as VPD increased, which is an
expected output given that both are a direct function of air temperature.
Figure 5.8. Canopy temperature deviation around solar noon (CTD) when fitted within the framework of Idso et al. (1981) and Idso (1982) during our experiments in 2013 and 2014. Each individual point is the CTD of each of 8 replicates per genotype, either Derrimut (pale red) or Gregory (pale blue). CTDUL and CTDLL represent, respectively, the CTD upper limit and CTD lower limit as explained in Figure 5.2.
To further analyse how CSWI behaved over time, we compared it with the trajectory of
the fraction of plant available water (FPAW) for each genotype and experiment. During
2013, CWSI started from close to zero (no water stress) at the beginning of October. As
the soil progressively dried, as shown by FPAW, the CWSI started to rise (Figure 5.9).
5.3 Results 163
The occurrence of rainfall on any given day abruptly reduced the CWSI, as shown, for
example, a couple of days before 21st October. The sharp reduction in CWSI during 17th
October was caused by cloudiness. It must be noted, however, that CWSI data during a
rainfall event or cloudy day is meaningless. The main aim for me to show this data in
Figure 5.9 is to verify the occurrence of the reduction of CWSI during rainfall or
cloudiness. Then, following with the CWSI data, there are unfortunately missing data
from 20th October to 7th November due to malfunction of the infrared thermometers'
data loggers—a misfortune considering that it must have been an interesting period to
look at CWSI when FPAW was at it lowest and leaf area, presumably, still fully green.
When data logging was recovered the 7th November, CWSI reached its maximum value
at around 0.7, which matched FPAW being at its minimum level of c. 0.3 during
October and November (Figure 5.9). Soon after the 76 mm of rainfall around 11th
November, say, the 14th November, the soil profile recovered FPAW to c. 0.6 which
corresponded to a reduction in CWSI to values around 0.5. Subsequently, rainfall events
during the rest of the growing season were not large enough as to increase FPAW
substantially. In part this was due to our inability to determine FPAW in the topsoil,
since our shallowest gypsum block was buried at 20 cm from the soil surface.
Therefore, these small rainfall events might have reduce the CWSI slightly (for
instance, the 20th November) but increases in FPAW were undetected because they
happened shallower than 20 cm depth. The same behaviour just described was observed
during 2014, where CWSI increased (more water deficit) as FPAW diminished over
time. This time the need to determine soil water content shallower than 20 cm was more
evident, since, for example, the 27 mm rainfall happening during 15th and 16th
164 Chapter 5
November reduced the CWSI from about 0.7 the 12th November to 0.5 the 18th
November, without a substantial increase in FPAW (Figure 5.10).
For both Figure 5.9 and Figure 5.10 it must be clarified that after the middle of the grain
filling period—say, from 19th November onwards—the increase in CWSI is most likely
caused by leaf senescence, that while reducing transpiration must have increased the
amount of net radiation that is partitioned to sensible rather than latent heat, increasing
canopy temperature. Unfortunately, we did not measure green leaf area index to validate
this assumption.
5.3 Results 165
Figure 5.9. Crop water-stress index (CWSI) calculated from canopy temperature (top), fraction of plant available water (FPAW) (middle), and rainfall per day (bottom) across time during 2013. Vertical, dashed lines across panels indicate anthesis (left) and physiological maturity (right) dates as in Figure 5.4. For CWSI and FPAW, points are the means, and bars the standard errors, of 8 replicates. For CWSI, data are missing between 20
th October and 7
th November due
to malfunction of the infrared thermometers' data loggers.
One might now wonder which is a better surrogate of the water status of a crop: CTD or
CWSI? To help answering this question we compared CTD and CWSI as functions of
FPAW (Figure 5.11). We excluded for this purpose data during periods when crop
senescence was high. Without having quantified green leaf area index, we deliberately
excluded data after mid grain-filling. It seems from Figure 5.11 that CWSI captures
more of the variation in FPAW than CTD alone. Both during 2013 and 2014, the R2 for
the relationships between CWSI and FPAW were higher than the R2 of the relationships
between CTD and FPAW (Figure 5.11). In particular, it is remarkable how, for the
relationship of CTD as a function of FPAW during 2013, the data points align when
166 Chapter 5
CTD is replaced by CWSI (Figure 5.11). Not so remarkable as 2013, but also during
2014 the data points align slightly better when the FPAW is expressed as a function of
CWSI rather than as a function of CTD. The biggest disagreement in the relationships
between CWSI and FPAW seem to happen at high values of FPAW (around 0.8), with
positive outliers during 2013 and several possible values of CWSI at the same level of
0.8 FPAW during 2014. This might happen because at this high values of soil available
water it is not the soil who limits transpiration but the demand from the atmosphere.
Although those likely changes in atmospheric demand for water must impact CWSI,
they are certainly independent of soil water status. Thereby, we can observe several
possible CWSI at the same level of (non-limiting) FPAW. This value of CWSI should
however be lower than 0.8, as the FPAW below which gas exchange is reduced is
around 0.4 (Sadras and Milroy 1996). This could be caused by our several sources of
uncertainty when using the framework of Idso (1982); namely, uncertainty when
calculating CTDLL from a model rather than measuring in a well-watered wheat plot,
and uncertainty in soil water content predicted by gypsum blocks.
5.3 Results 167
Figure 5.10. Crop water-stress index (CWSI) calculated from canopy temperature (top), fraction of plant available water (FPAW) (middle), and rainfall per day (bottom) across time during 2014. Vertical, dashed lines across panels indicate anthesis (left) and physiological maturity (right) dates as in Figure 5.4. For CWSI and PAW, points are the means, and bars the standard errors, of 8 replicates.
168 Chapter 5
Figure 5.11. A figure showing how crop water-stress index (CWSI) is a better predictor of the fraction of plant available water than the canopy temperature deviation from air temperature (CTD) alone. The figure shows either CTD (top row) or CSWI (bottom row) as a function of the fraction of soil available water (soil water held between suctions from 30 to 1500 kPa) during 2013 (left) or 2014 (right). Each data point is the mean of 8 replicates per genotype, for genotypes Derrimut (pale red) or Gregory (pale blue). R
2 are the r-squared values fitted per
genotype per year.
An increase in water use in a given day must reduce the CWSI on that day. To confirm
this we analysed how a difference in CWSI from one genotype to the other in any given
day impacted on their respective differences in water use on that day. To do that, we
took the data shown in Figure 5.11 (i.e. belonging to the first half of the grain-filling
period) and calculated each day the difference between the mean CWSI of Gregory and
the mean CWSI of Derrimut, and the difference in mean water use per day of Gregory
and the mean water use per day of Derrimut. The comparison is shown for both years in
Figure 5.12. The differences in CWSI between Gregory and Derrimut did not correlate
5.3 Results 169
with the differences in water use between them (Figure 5.12). This is striking, for a
higher water use, which equals transpiration under full canopy cover, must reduce
canopy temperature and therefore CWSI. Possibly, again, this is obscured by the large
level of uncertainties accumulated after every sequential step when calculating changes
in volumetric water content per day from soil water suction values coming from gypsum
blocks (Chapter 3). These uncertainties might have greater impact over a rate variable,
like water use per day, than over a state variable, like soil water content and therefore
FPAW—day to day changes might simply carry proportionally larger errors than soil
water content.
Nevertheless, most data fell within the quadrant where Gregory used more water and
had cooler canopy than Derrimut. Although there is no correlation, this is interesting
because it means that Gregory, the genotype that on average had less CWSI, also had
more water use per day (Figure 5.12).
170 Chapter 5
Figure 5.12. Crop water stress index (CWSI) of Gregory minus Derrimut, as a function of water use of Gregory minus Derrimut during 2013 and 2014. Axes are the values of Gregory minus the values of Derrimut—i.e. CWSI Gregory – CWSI Derrimut, and (water use Gregory) – (water use Derrimut). For instance, a value in the y-axis of -0.1 could mean that the CWSI of Gregory was 0.3 (less water-stressed) while the CWSI of Derrimut was 0.4 (more water-stressed). Correlation between ordinate and abscissa is nil, but data fall within the quadrant of negative difference in CWSI corresponding to positive difference in water use between Gregory and Derrimut—more water use corresponded with less water stress.
5.3.5 Crop growth and water use
Shoot biomass at maturity and grain yield were higher during 2013 than 2014 (Table
5.1). Gregory and Derrimut did not differ statistically in shoot biomass or grain yield,
though Gregory tended to produce more biomass (p = 0.06, Table 5.1). Moreover,
despite the lack of significance in the year × genotype interaction, the mean biomass
and yield from Gregory during 2013 and 2014 were higher than Derrimut (Table 5.1).
Therefore, both varieties produced more biomass during 2013 but the ranking between
them was maintained in both years: Gregory more than Derrimut. Harvest index was the
same within both varieties, and slightly higher during 2014 than 2013 (not significant,
5.3 Results 171
Table 5.1). Interestingly, total water use accumulated between anthesis and maturity
was higher during 2013 than 2014 (p < 0.05) and higher in Gregory than Derrimut by 11
mm (p < 0.05, Table 5.1). Interactions year × genotype were again non-significant but
followed, remarkably, the same ranking as biomass and yield, that is, Gregory 2013 >
Derrimut 2013 > Gregory 2014 > Derrimut 2014 (Table 5.1).
Table 5.1. Means of shoot biomass, grain yield, harvest index, and cumulative water use during grain filling, per year (top 3 rows), per genotype (middle 3 rows), and per year × genotype combination (bottom 5 rows). P values are specific to the rows above them. n.s. indicate comparisons that were not significant.
Year Genotype Shoot biomass (t ha
-1)
Grain yield (t ha
-1)
Harvest index (unitless)
Water use grain filling (mm)
2013 16.7 6.8 0.41 82.8
2014 12.1 5.1 0.42 63.3
P value < 0.0001 < 0.0001 n.s. < 0.05
Derrimut 13.0 5.4 0.42 66.6
Gregory 14.4 5.9 0.41 77.7
P value 0.06 n.s. n.s. < 0.05
2013 Derrimut 16.2 6.6 0.41 78.5
Gregory 17.3 7.1 0.41 88.8
2014 Derrimut 11.4 4.8 0.42 54.8
Gregory 12.9 5.3 0.41 70.7
P value n.s. n.s. n.s. n.s.
The consistent ranking in biomass, yield and water use during grain filling for each of
the year × genotype combinations suggest that they correlate with each other, and that
Gregory, while having extra deep water use (Figure 5.5 and Figure 5.6), was able to
uptake on average an extra 11 mm that translated into almost half a tonne of grain yield
172 Chapter 5
(Table 5.1). Indeed, at a plot to plot level, grain yield was related to water use during
grain filling with a slope of 54 kg ha-1 mm-1 (Figure 5.13).
Figure 5.13. Grain yield as a function of water used from anthesis to maturity for the varieties Derrimut and Gregory during 2013 (circles) and 2014 (triangles). Each data point is one replicate. The slope of the regression line, as well as its r-square and p-value, are shown.
5.4 Discussion
Two high yielding commercially grown spring wheat cultivars were chosen in this
study. Both are very competitive in the market place as they are high yielding in
Australian environments, have very good disease resistance, and have very acceptable
grain quality. Previous studies had detected subtle differences in them in that Gregory
had a reputation of extracting more soil water.
It was therefore expected that the techniques used in this study would need to be very
discriminating in order to detect differences in root systems, water extraction and
stomatal behaviour.
5.4 Discussion 173
Measurements made at maturity on both cultivars for grain yield, above-ground biomass
and harvest index were not significant at p = 0.05. The yield of Gregory was higher than
Derrimut but it just failed at p = 0.05. Above-ground biomass of Gregory was also
higher. Hence, this study was contrasting two very similar performing cultivars. The
finding that Gregory performed slightly better than Derrimut was consistent with all
aspects of this study, which measured root growth, canopy temperature and soil water
use.
5.4.1 Water use was higher in Gregory than Derrimut, but differences
in root length were not demonstrable
Gregory had higher root length densities beneath 1 m than Derrimut, but this could not
be shown to be statistically significant. Gregory also had higher water use in deep soil
layers than Derrimut (during 2013: p < 0.05). While these results are encouraging in the
sense that they suggest a correlation between deep roots and deep water use, alternative
explanations are possible. For instance, differences in deep water use could be a result
of a higher canopy conductance to water vapour in Gregory than Derrimut. Assuming
this alternative is true, higher deep water use in Gregory would be a result of a more
water-demanding shoot rather than a more efficient root system.
On the other hand, if the reverse is true, that is, if higher deep-water use is a result of
higher root length density in those layers in Gregory, then very small differences in root
length density must have produced differences in water uptake. The physiologically
relevant root length density for wheat has been reported at 0.1 cm cm-1 (Kirkegaard et
al. 2007). There is a degree of randomness in the sampling compounded with variability
174 Chapter 5
in root distribution in response to soil heterogeneity. Thus we cannot conclude with
confidence that the root length density values for Gregory fall beneath that threshold at
any of depths observed (to 180 cm). The same cannot be said for Derrimut, which does
appear to fall below that threshold at 170 cm in 2013 and 150 cm at the second site in
2014 (Figure 5.3).
In response to the greater water use of Gregory, yield increased 54 kg ha-1 mm-1, similar
water-use efficiencies that were estimated by simulation modelling by Manschadi et al.
(2006) and measured by Kirkegaard et al. (2007).
5.4.2 Canopy temperature was cooler in Gregory than Derrimut
Canopy temperature in Gregory was consistently lower than in Derrimut (Figure 5.4).
Given that water use was also typically higher in Gregory, this would seem to indicate
that, as expected, the canopy temperature result is linked to higher evapotranspiration.
However, the two measures did not correlate on a daily basis (discussed below) and the
fact that only 2 varieties were evaluated in this study limits our capacity to discern the
reasons for cooler canopy temperature. The cooler canopy temperature of Gregory could
also be due to other features of the canopy, such as canopy height, as found by Rebetzke
et al. (2012). Higher canopy height is supposed to reduce the resistance to the removal
of sensible heat by the wind. Thus, taller canopies dissipate heat quicker, and this heat
may be sensible (not latent) and as such unrelated to water use. Therefore, we cannot
assure that higher water use was the cause of a cooler canopy in Gregory than Derrimut.
Several genotypes would be needed in order to disentangle the contribution of water use
to canopy temperature by phenotyping in the way described in this study. I suggest that
5.4 Discussion 175
more genotypes should be evaluated together, so that canopy height and other above-
ground attributes confounding the relationship between canopy temperature and water-
use can be removed.
5.4.3 CWSI was a better surrogate of plant water-status than CTD
Using CWSI rather than CTD, or even more, canopy temperature per se, captured more
of the variation in plant available water (Figure 5.11). This happens because vapour
pressure deficit is one of the main drivers of transpiration, and by normalizing by it,
CWSI becomes an independent variable and has thereby more predictive value than
canopy temperature alone.
It is curious why such an easy to calculate variable is not used in breeding programs.
Perhaps it is because correct use would require determining the non-stressed baseline of
CTD (CTDLL) in a well-watered crop. This is not difficult if we consider that these
baselines might be species-specific but not varying between genotypes, as found in
maize (Cárcova et al. 1998). Once some plots are set up with irrigation, the baseline
functions relating CTDLL to VPD can be measured directly, and extrapolated to the
whole breeding trial if assuming they are equal in all plots. Then, the variables needed
in order to determine CWSI per plot are canopy temperature, air temperature, and air
relative humidity.
5.4.4 Daily changes in CWSI did not correlate with water use
Low canopy temperature has been presented as a surrogate for transpiration maintained
through access of roots to water deep in the soil profile (Reynolds et al. 2007; Lopes
176 Chapter 5
and Reynolds 2010; Pinto and Reynolds 2015). We replaced canopy temperature by
CWSI, which showed to be a more robust metric. However, although both canopy
temperature deviation (CTD) and CWSI were lower for Gregory, which used more
water than Derrimut, day-to-day changes in CWSI did not correlate with day-to-day
changes in water use. The reason could have been that the effect of the genotype over
water use was not large enough. Or perhaps they were caused by both errors
accumulated when calculating CTDLL from a model rather than measuring in a well-
watered wheat plot, or the uncertainty in soil water content predicted by gypsum blocks.
Despite CWSI was related to the fraction of plant available water (FPAW), this lack of
day-to-day correlation challenges the use of canopy temperature per se for phenotyping
wheat with higher water use and, incidentally, deep roots. Further conceptual models
are needed that can account for differences in canopy architectures between genotypes.
For example, the estimation of latent heat, and consequently evapotranspiration, by heat
balance as done by Choudhury et al. (1986) and Dunin et al. (1991), may be more
appropriate for phenotyping than canopy temperature per se. This is because heat
balance takes into account the height of the canopy for calculating the aerodynamic
resistance of the crop, and as such could potentially normalise the effect of canopy
height between Gregory and Derrimut. However, these two studies were done in large
field plots covered by a homogeneous wheat crop rather than as we did in the current
study, where of 12 m2-plots were arranged in a mosaic. Mixing canopies with different
heights in a mosaic may result in unpredictable aerodynamic resistances, and hence
highly uncertain estimations of evapotranspiration.
5.4 Discussion 177
5.4.5 Trial size is more important than continuous monitoring for
detecting overall differences between genotypes—but continuous
monitoring can reveal genotype interactions with time
One might wonder whether measuring canopy temperature continuously, as we have
done with the infrared thermometers fixed to each plot, is more convenient than
measuring in discrete times, e.g. 4 to 10 days during the whole season (Rebetzke et al.
2012; Pinto and Reynolds 2015). The cost of equipment, time and labour limits the
number of sensors that can be applied to plots in a trial, hence limiting the size of the
trial. However, once installed, the sensors can provide continuous monitoring of water
uptake over an extended period of time (barring data loss as experienced in 2013). Such
monitoring can provide information on genotype interactions with time but is limited in
its ability to detect genotype differences that are constant over time (genotypes main
effects) where the number of genotype replicates becomes more important. In cases
where having continuous monitoring of canopy temperature is not crucial,
measurements of canopy temperature by means of unmanned aerial vehicles equipped
with hyperspectral cameras may allow the sampling of a large number of plots.
Sampling many plots within a short period of time is important so that variations in
atmospheric conditions are minimised (e.g. VPD and solar radiation), which, as shown
in this Chapter can greatly influence canopy temperature.
Obviously, data must be collected over time in order to investigate genotype by time
interactions. Data collected over more time points also provides greater residual
degrees of freedom for the statistical detection of significant genotype by time
178 Chapter 5
interaction effects—although the improvement in statistical power experiences
diminishing returns as the residual degrees of freedom increase.
It should also be noted that observations made on the same plots at different time points
may exhibit strongly autocorrelated random behaviour, so that observations on
subsequent days do not represent independent observations of the plots over days—this
reduces the effective residual degrees of freedom gained from each subsequent days‘
observations (Greenhouse and Geisser 1959). For our 2014 trial, an autocorrelation of
approximately 0.65 was observed between successive days. Observations made over
larger time intervals (such as weekly intervals) are likely to be less strongly
autocorrelated.
5.4.6 Trial size also remains important for detecting smaller genotype
by time interactions overall
For detecting time-independent differences between genotypes (genotype main effects),
observations made over time have limited impact in improving accuracy compared to
trial size. In the relevant repeated measures analysis of variance of the data, the
significance of a genotype main effect is tested against the observed variance of the
replicated plots. Under the assumption of no autocorrelation between time points, the
observed variance is given by
⁄ , where is the true plot-to-plot variance
(the ‗plot variance component‘), and
⁄ is the contribution to the observed variance
from the random variation (of variance ) between n time points observed within plots.
5.5 Conclusions 179
Collecting data from more time points (increasing ) will only reduce the second term
above—the variance remains constant. The benefit to collecting more time points is
therefore limited by the relative sizes of and . As previously, the presence of
autocorrelated random behaviour will further reduce the improvement in accuracy (via
reduction of
⁄ ) obtained from each subsequent time point observed. In our 2014
dataset we observed no increase in the estimated standard error of the difference
between Gregory and Derrimut as we included more days of data into our analysis.
5.5 Conclusions
In summary, we showed that two high-yielding commercial wheat varieties seemed to
differ in deep root length, and had different canopy temperatures. Using a crop water-
stress index (CWSI), rather than canopy temperature per se or canopy temperature
deviation from air temperature, gave better predictions of plant water-status—at least
concerning the amount of plant available water present along the soil profile. The CWSI
of the variety with deep roots, Gregory, was lower than the one with shallower roots,
Derrimut. Gregory also used more water, but the difference in water use per day
between both varieties did not correlate with their difference in canopy temperature.
Extra water use in Gregory came from deeper in the soil, and was used during grain
filling to grow more grain at a rate of 54 kg ha-1 mm-1. When compared to average
values of transpiration efficiency from sowing to maturity (c. 24 kg ha-1 mm-1), the yield
produced per unit of water used during grain filling is doubled. This indicates that deep
180 Chapter 5
roots can be detected by CWSI and can increase yield by taking advantage of water
during grain filling when it is most efficiently used.
181
Chapter 6 General discussion and implications
for crop improvement
6.1 Overview of the thesis
The genetic improvement in rooting depth is a major challenge to the on-going effort to
improve the yield of crops, particularly those grown in water-limited conditions. The
selection for root system traits, especially deep roots, remains challenging because of
the difficulties involved in root sampling in field conditions. We also know very little
about what to select for when it comes to root systems and how to select for them. In an
attempt to help in the solution of these problems, this study aimed to determine whether
genetic diversity in rooting depth at maturity existed between wheat and triticale
genotypes growing in field conditions. We found that genotypes differed in rooting
depth (p < 0.10). Differences were larger, however, between species (triticale and
wheat) than within a species (wheat). Within a population of wheat genotypes that
express a gradient of plant heights, genotypes differed in rooting depth (p < 0.10),
biomass (p < 0.10) and yield (p < 0.05). Interestingly, the shorter the plants, the deeper
the roots and higher the yield, which suggest that a shift in carbon partitioning from
stems towards grain and roots had happened.
These differences in rooting depth, while encouraging, were obtained by soil coring, a
very time-consuming process that is difficult to adopt in breeding programs. Thus, our
second aim was to investigate whether canopy temperature can be used as a surrogate of
182 Chapter 6
transpiration, and thereby as a predictor of deep water use. Canopy temperature would
then be used to select wheat genotypes with deep roots. For this, we compared canopy
temperature and water use between Gregory, a wheat with tentatively deeper roots, and
Derrimut, a wheat with shallower roots. Both varieties are high-yielding commercial
wheats sown in Australia. Canopy temperature reflects an energy balance where many
factors are involved, so we calculated a water-stress index that normalized for the main
factors affecting the relationship between canopy temperature and transpiration: air
temperature and vapour pressure deficit. We found that genotypes did not differ in
rooting depth or deep root length, but mean values of these root traits were higher in
Gregory than Derrimut in three independent experiments. Indeed, Gregory had deeper
water use (p < 0.05) and more total water use during grain filling (p < 0.05) than
Derrimut, and it also had a cooler canopy during that period which corresponded to a
lower crop water-stress index. Finally, the crop-water stress index correlated closely
with the amount of soil available water, confirming that this index calculated from
canopy temperature could be used to select wheat with deeper water use.
In the following sections I will outline the main results of the thesis, (i) highlighting the
importance of deep roots for wheat yield, (ii) listing ways in which breeding can
increase rooting depth, and (iii) proposing future work that would be interesting to
undertake following this thesis.
6.2 The benefits of deeper roots to wheat growth and yield
One of the most significant findings to emerge from this thesis is that rooting depth
differed between wheat and triticale genotypes growing in field conditions (p < 0.10).
6.2 The benefits of deeper roots to wheat growth and yield 183
To my knowledge, there are no studies reporting genotypic differences in rooting depth
other than this thesis and the study by Wasson et al. (2014). Botwright Acuña and Wade
(2012) evaluated rooting depth in the field in 24 wheat cultivars growing at 6 locations
in Western Australia. They found that the sum of squares given by genotype ×
environment interactions was more than 3 times larger than the sum of squares given by
genotype, but the latter was not statistically significant. Others have suggested
differences in rooting depth and deep root length by inferring them from water use deep
in the soil. For instance, Pask and Reynolds (2013) measured water uptake by roots at
several soil depths under non-limiting moisture conditions in a set of wheat varieties
released between 1950 to 2009. They found that modern wheat varieties take up more
water from deep soil than older varieties. There was an increase in water uptake
between 60 and 120 cm, but most markedly between 90 and 120 cm. Interestingly, this
increase by means of breeding was responsible for gains in yield potential (that is, yield
only limited by radiation and temperature, not water or nutrients). However, despite
their results being encouraging, these authors measured neither root length density nor
rooting depth, as we did in this thesis. Although they suggest that deeper water use was
caused by more root length at depth, deeper water use can also be due to an inherently
higher stomatal conductance in the modern varieties. In the US corn-belt, Hammer et al.
(2009) suggested that rooting depth in maize was changed by breeding, although this
was not measured directly but rather it was inferred by simulation modelling. The
results of this thesis are one of the few showing rooting depth differences among
genotypes in the field.
184 Chapter 6
What is also interesting is that those differences in rooting depth and deep root length
correlated with shoot biomass and grain yield. Now, it is true that the main source of
variation in the data in Figure 4.3 was soil type and the amount of rainfall in each
season, with less variation due to genotype. But as shown in Figure 4.4, there was one
experiment where shoot biomass, grain yield and even harvest index increased as the
root length below 120 cm increased. Interestingly, this experiment was carried out
during 2012, when soil available water was not apparently limiting. Pask and Reynolds
(2013) also found that biomass and yield correlated with greater water use between 90
and 120 cm depth, and Lopes and Reynolds (2010) found that biomass of deep roots
was related to the amount of residual water left at maturity—the greater the amount of
deep-root biomass the lesser the amount of residual soil water. Although Pask and
Reynolds (2013) and Lopes and Reynolds (2010) did not measure root length density,
their results, and the results of this study, highlight the importance of deep water use
even in conditions of mild water deficit.
Another interesting finding of this thesis is that, by having deeper roots, genotypes
increased total water use during the season, and yield at with a high efficiency (Chapter
5). Though one must be cautious given that this is inferred from two experiments where
only two genotypes were evaluated, the results imply that having deep roots is highly
valuable in terms of water economy for wheat. The rate of grain yield per unit of water
used during grain filling was 54 kg ha-1 mm-1, more than double the usual 22 kg ha-1
mm-1 reported for the whole duration of the growing season (French and Schultz, 1984;
updated to 24 kg ha-1 mm-1 by Sadras and Lawson, 2013). This result of higher water-
use efficiency during grain filling is similar to results published by Manschadi et al.
6.3 Ways of improving root depth by breeding 185
(2006) and Kirkegaard et al. (2007). The fact that water-use efficiency to produce grain
is higher during grain filling can be explained by the fact that (i) during that period it is
mostly grains that are growing (no vegetative organs), and (ii), as shown by Borrás et
al. (2004), crops in general (though maize more so than wheat) respond markedly to
improvements in growing conditions during grain filling if assimilates for grain growth
are limiting. This is because the potential capacity for grain growth is set early in grain
filling, so a crop that established a high potential grain size early would respond
significantly to alleviations of stresses later during grain filling (Borrás et al. 2004).
The current study also suggests that the benefits of deep roots are not exclusive to
conditions where drought prevails during early reproductive development or late grain
filling (terminal drought). On the contrary, they are still meaningful in an environment
with mild or no water stress. This was shown by the range of yields covered in Chapter
4 and 5 (from ~ 3 to 8 t ha-1). But this can be seen particularly in the case of experiment
3, where biomass, yield and harvest index responded to root length density under 120
cm depth while yields ranged between 5 and 8 t ha-1. That yield increases with deeper
roots or deeper water use is consistent with the results of Pask and Reynolds (2013).
6.3 Ways of improving root depth by breeding
A key strength of the present thesis was the finding that rooting depth can vary with
germplasm. It was usually documented in controlled conditions (Manschadi et al. 2006,
2008), but there are a few cases where rooting depth is measured and found to differ
between genotypes under field conditions.
186 Chapter 6
A question that emerges is how can breeding increase rooting depth. A reasonable
approach to tackle this issue would be to focus on the traits that allowed for greater root
growth. Examples of traits to which we engaged in this thesis, as mentioned in Chapter
1, are (i) early vigour, (ii) reduced tillering, (iii) reduced height, as well as, (iv) deeper
rooting directly. To tackle these traits and their effects on root growth, we could first
classify the genotypic differences that we found in this thesis as interspecific, i.e.
between different species (triticale and wheat), or intraspecific, i.e. between genotypes
of the same species (wheat differing in height). The distinction is important because to
breed wheat by introgressing genes from a different species is far more difficult than
introgressing from another wheat genotype.
Concerning interspecific differences, triticale produced deeper roots, more biomass and
higher yield than wheat (p < 0.10). These results are interesting because they confirm
the higher resilience of triticale to drought found by Giunta et al. (1993) and Condon
(personal communication). However, the application of this knowledge to wheat
breeding might be limited considering that introgressing genes from triticale to wheat
may prove a challenge. But as proposed in the preceding paragraph, one could better
investigate what is the trait that enabled triticale to have such a great advantage in terms
of root and shoot growth than wheat. The most evident one seems to be the fact that
triticale is just more vigorous early in seedling development, and later those differences
accumulate towards a mature plant, which result eventually in higher shoot growth than
wheat. Apparently, that above-ground vigour is mirrored by high vigour in the roots as
well. If this is the case, then this challenge of increasing early vigour in wheat was
already undertaken by Rebetzke and Richards (1999).
6.3 Ways of improving root depth by breeding 187
Concerning our intraspecific differences, we found that, for wheats differing in height,
shorter stems were related with deeper root growth and more yield. It seems, therefore,
that alleviating the competition for assimilates between stems and roots increased
rooting depth. Taken together, these findings support the concept of the ideotypes and
communal plants that was originally drawn by Donald (1968). This is interesting since
the change in roots during the Green Revolution is not known. In young wheat
seedlings, Wojciechowski et al. (2009) showed that Rht genes had an effect on root
growth without an effect in shoot growth, suggesting that the effect of Rht genes on
roots is direct. Richards (1992) and Miralles et al. (1997), on the other hand, suggested
that the effect of these genes was not direct but rather secondary, i.e. by reducing shoot
growth more assimilate was available for root growth. In accordance with Richards
(1992) and Miralles et al. (1997), shoot biomass was reduced as plant height decreased.
Miralles et al. (1997) also found that, while plant height was reduced, total root length
and root dry weight were increased. They suggested that this was due to a surplus of
assimilates without a sink to utilise them. Miralles et al. (1997) found no difference in
root length but rather in root dry weight, that they propose is a result of roots thickening
but not elongating. Returning to the results of this thesis, however, these data related to
the increase in rooting depth as plant height increased must be interpreted with caution
because regressions in Figure 4.5 where only significant if considering p < 0.10.
Reduced tillering has been proposed as a way of increasing yield when crops rely on
stored soil moisture (Hendriks et al. 2015). In this thesis, however, we were unable to
detect differences in root growth, shoot biomass or yield between isogenic lines for the
tin gene. Perhaps this was because our crops had more available water in-season. The
188 Chapter 6
main advantage of tin genes in wheat seems to be in crops that rely on in-season
rainfall. In this context, while leaf area expansion is reduced in tin isogenic lines, so too
was the relative amount of rainfall lost by direct evaporation. This trait would be
detrimental in the opposite growing conditions, when crops rely heavily in stored soil
moisture. Under these conditions, faster leaf area expansion would lead to a greater
transpiration before anthesis and a consequent reduction of the amount of water
available for grain filling. This in turn would reduce harvest index and grain yield
(Passioura 1976). In our conditions, crops did not experience terminal drought during
2011 and 2012, and therefore tin lines provided no advantage over lines without tin.
Liao et al. (2006) found that vigorous lines produced more roots from 20 to 70 cm depth
in glass-walled pots, with no differences in rooting depth. Watt et al. (2005) found
faster root growth in the more vigorous line V18 than the non-vigorous genotype Janz,
when sampled at 2 and 5 fully expanded leaf stages. However, the differences declined
towards anthesis and maturity (Watt et al. 2013). Therefore, early vigour might be
beneficial in the same sense as reduced tillering: increasing the fraction of total
evapotranspiration that is transpired and reducing direct soil evaporation in
environments where wheat relies on in-season rainfall (Palta et al. 2011). In a field
study during two seasons, Botwright Acuña et al. (2002) documented that early vigour
increased yield by c. 12% in the wetter year, but there were no effects in the drier year.
In a simulation study evaluating increased early vigour by means of increasing specific
leaf area (SLA, cm2 g-1) in APSIM-Nwheat, Asseng et al. (2003) found that increased
SLA only increased yields in sandy soils, adding to the notion that early vigour might
augment transpiration while reducing direct evaporation from the soil. Therefore, the
6.3 Ways of improving root depth by breeding 189
contribution of early vigour to greater rooting depth and deep water uptake is uncertain
from the start.
Another trait that might increase the potential of wheat to access deep water is the
capacity of the root system to penetrate hardpans. Botwright Acuña and Wade (2005)
tested in soil columns the ability of wheat roots from several genotypes to penetrate thin
wax layers. These wax layers were used to simulate soil hardness. They found
genotypic differences in the capacity of roots to grow through the wax layers.
Moreover, Botwright Acuña et al. (2007) showed that those genotypes with more root
length passing through the wax layers were the same genotypes exhibiting the highest
rooting depths in the field. This indicates that in some environments the ability of some
genotypes to grow roots through soil layers with high mechanical impedance can
improve rooting depth and possibly deep water use.
It is important to mention, however, that it is worth defining the target environment
where the trait would be of any advantage for grain yield, for it would be useless if it
could improve deep root growth but have a detrimental effect over yield. This is
provided by a pivotal analysis done by Chenu et al. (2013) for wheat grown in the
whole Australian wheat-belt. Chenu et al. (2013) divided the whole area of the
Australian wheat-belt into four environmental types. This is interesting because the
environments, rather than being defined solely by their climate and soil conditions, are
defined from 'the point of view of a crop facing drought'. This 'point of view' is defined
as an index of supply over demand, where supply is calculated as the amount of
available water within the rooting depth, and demand is calculated from crop growth
rate and the transpiration efficiency (which is species-dependent and normalized by
190 Chapter 6
vapour pressure deficit [the higher the vapour pressure deficit the lower the transpiration
efficiency]). The analysis is summarized in Figure 6.1, where, based on the most
common patterns of source-demand ratio along the development of the crop (with a
centre at anthesis), they defined the four environment types. Those environments were:
environment 1: no or light stress;
environment 2: post-flowering mild stress;
environment 3: stresses that began before flowering but were relieved during grain
filling; and
environment 4: severe stresses that began during the vegetative period and lasted
until maturity.
6.3 Ways of improving root depth by breeding 191
Figure 6.1. Simulated water supply-demand ratio for four environment types that typify the most common water balances for wheat across the Australian wheat-belt as a function of thermal time accumulated from anthesis (base temperature = 0 °C). Numbers indicate 4 possible environment types: environment 1: no or light stress; environment 2: post-flowering mild stress; environment 3: stresses that began before flowering but were relieved during grain filling; and environment 4: severe stresses that began during the vegetative period and lasted until maturity. Redrawn from Chenu et al. (2013 fig. 5).
Depending on the environment type where a putative wheat genotype expressing the
trait under consideration is grown, the effect on yield will be positive or negative. For
instance, a genotype with increased early vigour would only suit environment 1, where
a faster leaf area expansion early in the season would minimize water wasted by direct
soil evaporation. In this environment, continuous (and perhaps small) rainfall events
would guarantee access to water in shallow soil layers with little risk of running out of
water for late grain filling. The opposite could be expected for a tin wheat genotype
with reduced tillering, where limited tillers early in the season would slow down the
increase in leaf area index and increase direct soil evaporation. However, in the later
192 Chapter 6
part of the season, there would be more soil water to spend during the reproductive
period. A reduced tillering genotype would take advantage of environments 2, 3 and 4.
Similarly, a genotype possessing deep roots would be advantageous in these three
environment types.
Finally, if traits are genetically independent from each other (i.e. there is no linkage
between chromosome regions on which they depend) then they could in principle be
combined in the same plant, they can be pyramided (Wilson et al. 2015). Therefore,
ideotypes can be targeted for each environment type: presence of in-season rainfall
(environment 1) = high early vigour, rapid canopy closure, vigorous shallow roots;
reliance on stored soil water (environments 2, 3, 4) = reduced tillering, high carbon
isotope discrimination (i.e. high transpiration efficiency), and deep roots.
6.4 Phenotyping for deep roots
Continuously measuring canopy temperature proved useful, for it enabled us to
calculate the crop water stress index (CWSI) over time and predict soil water
availability. In other studies canopy temperature has been measured, roughly, every 3-
10 days by operators with hand-held infrared thermometers (Lopes and Reynolds 2010;
Rebetzke et al. 2012; Pinto and Reynolds 2015). By measuring canopy temperature
continuously, we were able to select days when canopy temperature is more
representative of transpiration, like clear days with low wind speed. While this can be
done by sampling weekly and choosing days with those qualities, investing in devices
that allow for continuous sampling could guarantee more days of sampling and perhaps
less cost. Moreover, it is somewhat surprising that CWSI is not so widespread in canopy
6.4 Phenotyping for deep roots 193
temperature studies. It would be easy to incorporate CWSI as a selection criteria rather
than canopy temperature or canopy-air temperature deviation. All that is required is to
know relative humidity of the air, which can be easily obtained.
We found that cool canopy temperature and a low CWSI were related to higher water
use which in turn was directly related to grain yield. However, caution is needed in
interpreting these results, given that we only evaluated two genotypes. Although other
interesting studies have evaluated just two varieties (e.g. Condon et al. (2002)), other
factors that differed between those two varieties like biomass or plant height (Rebetzke
et al. 2012) could influence their canopy temperature and its relationship to
transpiration. It is crucial to compare genotypes with similar time to flowering, since
differences in flowering time would put green leaf area indices out of phase between
varieties, some senescing earlier, and some senescing later. Matching green leaf areas
would help to eliminate the possible effects of a diminishing leaf area index on
transpiration due to senescence. In Condon et al. (2002) both time to flowering and
plant height were equal between the two genotypes evaluated, Quarrion and Matong. In
our study, Derrimut and Gregory differed in plant height, the latter being 20 cm taller.
Moreover, these genotypes were some of the top yielding varieties in Australia, and
therefore few differences were expected between them. It is suggested that more
genotypes should be evaluated together to test whether CWSI offers potential as a
predictor of water use and deep root length.
Gypsum blocks were not accurate for determining volumetric soil water content, but
they could be employed for phenotyping soil water suction. If one intends to measure
water use precisely to relate to canopy temperature or CWSI, the best alternative would
194 Chapter 6
be to use lysimeters. However, this is impractical for selecting between hundreds of
genotypes within a breeding program. Therefore, gypsum blocks, in spite of being
inaccurate compared with lysimeters, can predict soil water use for many plots, and be
used for phenotyping wheat. Errors can be reduced by sowing trials in homogeneous
paddocks, or avoiding the need of calculating volumetric water content by just looking
at the changes of soil water suction as a way of detecting genotypes with access to deep
water. The limits for soil water suction within which plants can take up water was
within the measurement range of the gypsum blocks. Gypsum blocks could not reach
soil water suctions close to 30 kPa, as is usually assumed for field capacity, but they
predicted well between 50 kPa and close to wilting point, at 1500 kPa.
6.5 Future research
6.5.1 Automation of soil coring
One of the questions raised by this thesis is whether sampling roots by soil coring is
worthwhile if practiced in the way we did. This involved counting roots by a naked eye
and several operators. The fact that root length density was so variable after so much
time-consuming sampling might indicate that this is not the best way to phenotype root
systems. How roots spread along the soil volume is erratic, as it is strongly influenced
by local changes in soil hardness, moisture, and nutrients. But another part of the
variation is strictly experimental error when sampling. It could be possible to drastically
reduce that experimental error by removing human error by using automated ways of
counting. A fluorescence imaging system with root counting software has been
6.5 Future research 195
developed for this purpose (Wasson, personal communication). This is promising since
the errors in root counting and root length density determination could be reduced and
the sampling made more cost effective by reducing the labour requirements.
6.5.2 Measuring root function without measuring roots
Sampling roots is both time consuming and the parameter values are variable. One
could be radical and avoid sampling roots altogether. Instead, we could measure root
activity rather than roots per se. This concept is based on the approach of the pioneer
crop modellers De Wit and Penning de Vries (De Wit and Penning de Vries 1983), who
proposed 'modelling plant hormone action without modelling the hormones' (Hammer et
al. 2004). By following these criteria, we could measure root water uptake (i.e. root
function) without measuring root length density, inferring, indirectly, root presence
from root function.
This was done successfully by Dardanelli et al. (1997), who used a model developed by
Passioura (1983) to formalise root water uptake from a given soil layer in the same
fashion as a soil dries
(6.1)
and therefore
(6.2)
where is available soil water content over the wilting point (m3 m-3) at any given time
(days); is soil available water content at the start of soil water extraction (m3 m-3);
is the threshold time at which water extraction starts; is time since the
196 Chapter 6
beginning of water extraction; l is root length density (cm cm-3); and k is a parameter
with the dimensions of soil water diffusivity (cm2 d-1) but actually summarizes an
apparent diffusivity of the root-soil combined. It must be noted that, in spite of this
parameter k having the same dimensions as soil water diffusivity, it is not soil water
diffusivity, which is in turn orders of magnitude larger than k (Deery 2008; Deery et al.
2012).
In this way, kl summarizes the capacity of the root system to capture water in a given
soil layer per day. It is the maximum water uptake that can be achieved in one day, and
is a time constant, or in other words a kl of 0.10 d-1 means that today's water uptake is
10% of soil water left in the soil since yesterday, and tomorrow it will be the 10% of
what is left after in the soil today. It decays exponentially. The parameter kl is indeed
applied by the widely used simulation model Apsim-Nwheat to simulate root water
uptake (Keating et al. 2001).
The process is shown in Figure 6.2, where it can be seen on the left panel how kl is
calculated per soil depth, and on the right panel how kl distributes along the soil profile.
What Dardanelli et al. (1997) found is that a sunflower commercial hybrid called
Contiflor 3 had consistently higher kl than another hybrid called G100. This means that
the former hybrid had a root system able to capture more water than the latter, although,
because of the inherent properties of the method, it could not be disentangled whether
this was due to more root length density or more root hydraulic conductivity.
For kl to be the maximum possible water uptake per day it is necessary that the demand
of water by the atmosphere is always greater than the capacity of the soil, interacting
6.5 Future research 197
with the roots, to supply it. This was discussed in Chapter 1 where the concept from
Cowan (1965) was presented, in which transpiration will be equal to the supply or equal
to the demand, whichever is smaller. In other words, the only way of isolating the
capacity of roots and soil to provide water for transpiration is by guaranteeing that
demand is always larger than supply. This is a drawback because it means that this
study needs to be done either under a rain-out shelter or in arid environments.
Figure 6.2. Depiction of the method for determining the maximum water uptake per day per soil
layer (kl). On the left, the graph shows the available soil water content at a given time t ( ) as a function of time since sowing for three soil depths. The parameters of equations (6.1) and (6.2) are indicated as well. On the right, the kl calculated as in the graph in the left are plotted along soil depth for two sunflower hybrids. Bars are standard errors. It must be noted that the data on the left is peanut and therefore does not relate to the data on the right; it only serves as an example. The figure was redrawn from Dardanelli et al. (1997).
The rewards may make this technique worthwhile, since, to my knowledge, Dardanelli
et al. (1997) represents the only study carried out in field conditions where differences
in root capacity to capture water were detected between genotypes.
198 Chapter 6
6.5.3 Enhanced root vertical distribution, trade-off-free opportunities
and root self/non-self recognition
As previously discussed, a shift of root biomass allocation from shallow to deep soil
layers could be more effective for capturing soil water while minimizing the
expenditure of assimilates in extra root growth. This shift might already have happened
in maize hybrids progressively released in the US corn-belt (Hammer et al. 2009).
As suggested by Denison (2012, 2015), breeding should tackle trade-off-free
opportunities, i.e. opportunities that were rejected by past natural selection because they
were detrimental to individual plants, but are nowadays favourable to agriculture
because they benefit the plant population forming a crop as a whole.
Perhaps the key to shifting root growth to deep soil is to investigate into how root tips
perceive other root tips: as self or non-self? Self roots would not compete with each
other because they would reduce the fitness of their mother plant and the genes causing
it, being completely shared by both roots, would be eradicated by natural selection.
Non-self roots, belonging to different plants, might have different genes, and by
competing with each other for water or nutrients they might have increased their own
fitness and past natural selection have favoured them. Therefore, recognizing whether
another root is self or non self must offer advantages, and indeed this sort of recognition
exists (Gersani et al. 2001; Falik et al. 2003).
The most common response to encounters between root tips that belong to different
mother plants is branching ending up in root overproduction (Gersani et al. 2001;
Depuydt 2014). This might explain in part why root length density is usually higher in
6.6 Conclusion 199
the topsoil than the 0.1 cm cm-3 (Deery 2008) that is required to maximize water uptake.
Thus, it would be useful for plants forming a wheat crop to avoid overproduction of
roots in the topsoil by inhibiting the mechanism that enables non-self recognition. But it
may not be just a matter of genetic differences between confronting roots, as Gruntman
and Novoplansky (2004) found that roots coming from cuttings from the same
individual of the perennial grass Buchloe dactyloides, were progressively behaving as
non-self as the time since the cutting progressed. This suggest that the recognition is not
genetic but physiological, possibly mediated by root exudates (Semchenko et al. 2014).
In either case, the suppression of the non-self root recognition behaviour from crop
plants may make them more effective in using water and nutrients by occupying the soil
more effectively and relieving extra carbon to enable root growth deep in the profile.
6.6 Conclusion
The unifying hypotheses in this study were that (i) there is genotypic variation in the
depth of rooting that can be used to increase growth of the crops and grain yield; and
that (ii) canopy temperature and soil water measured continuously can be used to
phenotype wheat with deeper water access and incidentally deep roots. Genotypic
variation for rooting depth, both interspecific (triticale and wheat) and intraspecific
(variable height lines of wheat), were detected in field conditions. Deep rooting was
shown to be positively correlated with biomass and yield, and to be associated with
increased water uptake and a higher water-use efficiency. Gypsum blocks were not a
suitable tool to phenotype water uptake, but may be appropriate where suction is being
measured. Continuous measurement of canopy temperature was shown to allow
200 Chapter 6
phenotyping of root systems with superior deep water access, by enabling the
calculation of a CWSI that would be challenging to achieve with sporadic
measurements. Taken together, these findings can help breeders to produce wheat
varieties that can make better use of deep water and improve yield in water-limited
environments.
201
Appendix A ASReml-R models
This section presents specific code written in the R language for data analysis (R Core
Team 2014).
Model (4.1):
asreml(fixed = biomass ~ genotype * site + at(site, 2):lin(row),
random = ~ at(site):block,
rcov = ~ at(site, 1):column:ar1(row) +
at(experiment, 2):ar1(column):row,
data = data,
na.method.X = "include")
Model (4.2):
asreml(fixed = biomass ~ genotype + lin(column) + lin(row),
random = ~ block,
rcov = ~ column:row,
data = data,
na.method.X = "include")
Model (4.3):
asreml(fixed = biomass ~ group * site + at(site, 2):lin(row),
random = ~ at(site):block + genotype,
rcov = ~ at(site):ar1(column):row,
data = data,
na.method.X = "include")
203
Appendix B Validation of gypsum blocks
calibration against neutron probe
Figure B.1. Comparison of the soil water content estimated by gypsum blocks (lines) with the soil water content estimated with neutron probe (dots) in the soil layer corresponding to 20 cm depth. The red line is the trajectory of soil water content estimated with gypsum blocks fitted to the water retention curve calibrated in the pressure plate. The green line is the trajectory estimated with the water retention curve estimated by fitting gypsum block suction to soil water content given by neutron probe. Each frame is a plot from experiment 4 (2013), where column numbers are 'columns' and row numbers are 'rows' of that experiment (see Figure 2.11 for a detailed map).
204 Appendix B
Figure B.2. Comparison of the soil water content estimated by gypsum blocks (lines) with the soil water content estimated with neutron probe (dots) in the soil layer corresponding to 40 cm depth. The red line is the trajectory of soil water content estimated with gypsum blocks fitted to the water retention curve calibrated in the pressure plate. The green line is the trajectory estimated with the water retention curve estimated by fitting gypsum block suction to soil water content given by neutron probe. Each frame is a plot from experiment 4 (2013), where column numbers are 'columns' and row numbers are 'rows' of that experiment (see Figure 2.11 for a detailed map).
Appendix B 205
Figure B.3. Comparison of the soil water content estimated by gypsum blocks (lines) with the soil water content estimated with neutron probe (dots) in the soil layer corresponding to 60 cm depth. The red line is the trajectory of soil water content estimated with gypsum blocks fitted to the water retention curve calibrated in the pressure plate. The green line is the trajectory estimated with the water retention curve estimated by fitting gypsum block suction to soil water content given by neutron probe. Each frame is a plot from experiment 4 (2013), where column numbers are 'columns' and row numbers are 'rows' of that experiment (see Figure 2.11 for a detailed map).
206 Appendix B
Figure B.4. Comparison of the soil water content estimated by gypsum blocks (lines) with the soil water content estimated with neutron probe (dots) in the soil layer corresponding to 80 cm depth. The red line is the trajectory of soil water content estimated with gypsum blocks fitted to the water retention curve calibrated in the pressure plate. The green line is the trajectory estimated with the water retention curve estimated by fitting gypsum block suction to soil water content given by neutron probe. Each frame is a plot from experiment 4 (2013), where column numbers are 'columns' and row numbers are 'rows' of that experiment (see Figure 2.11 for a detailed map).
Appendix B 207
Figure B.5. Comparison of the soil water content estimated by gypsum blocks (lines) with the soil water content estimated with neutron probe (dots) in the soil layer corresponding to 100 cm depth. The red line is the trajectory of soil water content estimated with gypsum blocks fitted to the water retention curve calibrated in the pressure plate. The green line is the trajectory estimated with the water retention curve estimated by fitting gypsum block suction to soil water content given by neutron probe. Each frame is a plot from experiment 4 (2013), where column numbers are 'columns' and row numbers are 'rows' of that experiment (see Figure 2.11 for a detailed map).
208 Appendix B
Figure B.6. Comparison of the soil water content estimated by gypsum blocks (lines) with the soil water content estimated with neutron probe (dots) in the soil layer corresponding to 120 cm depth. The red line is the trajectory of soil water content estimated with gypsum blocks fitted to the water retention curve calibrated in the pressure plate. The green line is the trajectory estimated with the water retention curve estimated by fitting gypsum block suction to soil water content given by neutron probe. Each frame is a plot from experiment 4 (2013), where column numbers are 'columns' and row numbers are 'rows' of that experiment (see Figure 2.11 for a detailed map).
209
References
Aitchison G, Butler P (1951) Gypsum block moisture meters as instruments for the
measurement of tension in soil water. Aust J Appl Sci 2, 257–266.
Amani I, Fischer RA, Reynolds MP (1996) Canopy temperature depression association
with yield of irrigated spring wheat cultivars in a hot climate. Journal of Agronomy and Crop Science 176, 119–129. doi:10.1111/j.1439-037X.1996.tb00454.x.
Asseng S, Turner NC, Botwright T, Condon AG (2003) Evaluating the impact of a trait for increased specific leaf area on wheat yields using a crop simulation model.
Agronomy Journal 95, 10–19.
Blum A (2009) Effective use of water (EUW) and not water-use efficiency (WUE) is the target of crop yield improvement under drought stress. Field Crops Research
112, 119–123. doi:10.1016/j.fcr.2009.03.009.
Borrás L, Slafer GA, Otegui ME (2004) Seed dry weight response to source–sink
manipulations in wheat, maize and soybean: a quantitative reappraisal. Field Crops Research 86, 131–146.
Borrell AK, Hammer GL, Douglas ACL (2000) Does maintaining green leaf area in
sorghum improve yield under drought? I. leaf growth and senescence. Crop Science 40, 1026–1037. doi:10.2135/cropsci2000.4041026x.
Botwright Acuña TL, Condon AG, Rebetzke GJ, Richards RA (2002) Field evaluation of early vigour for genetic improvement of grain yield in wheat. Crop and Pasture Science 53, 1137–1145.
Botwright Acuña TL, Pasuquin E, Wade LJ (2007) Genotypic differences in root penetration ability of wheat through thin wax layers in contrasting water regimes
and in the field. Plant and Soil 301, 135–149. doi:10.1007/s11104-007-9428-9.
Botwright Acuña TL, Wade LJ (2005) Root penetration ability of wheat through thin wax-layers under drought and well-watered conditions. Australian Journal of
Agricultural Research 56, 1235–1244.
Botwright Acuña TL, Wade LJ (2012) Genotype × environment interactions for root
depth of wheat. Field Crops Research 137, 117–125. doi:10.1016/j.fcr.2012.08.004.
Bouyoucos GJ, Mick AH (1940) An electrical resistance method for the continuous
measurement of soil moisture under field conditions. Technical Bulletin Michigan Agricultural Experiment Station 172, 38 pp.
Butler D (2009) ‗asreml: asreml() fits the linear mixed model.‘ www.vsni.co.uk.
210 References
Campos H, Cooper M, Habben JE, Edmeades GO, Schussler JR (2004) Improving drought tolerance in maize: a view from industry. Field Crops Research 90, 19–34. doi:10.1016/j.fcr.2004.07.003.
Cárcova J, Maddonni GA, Ghersa CM (1998) Crop water stress index of three maize hybrids grown in soils with different quality. Field Crops Research 55, 165–
174. doi:10.1016/S0378-4290(97)00076-2.
Chandler PM, Harding CA (2013) ‗Overgrowth‘ mutants in barley and wheat: new alleles and phenotypes of the ‗Green Revolution‘ Della gene. Journal of
Experimental Botany 64, 1603–1613. doi:10.1093/jxb/ert022.
Chandler PM, Marion-Poll A, Ellis M, Gubler F (2002) Mutants at the slender1 locus of
barley cv himalaya. molecular and physiological characterization. Plant Physiology 129, 181–190. doi:10.1104/pp.010917.
Chenu K, Deihimfard R, Chapman SC (2013) Large-scale characterization of drought
pattern: a continent-wide modelling approach applied to the Australian wheatbelt – spatial and temporal trends. New Phytologist 198, 801–820.
doi:10.1111/nph.12192.
Choudhury BJ, Reginato RJ, Idso SB (1986) An analysis of infrared temperature observations over wheat and calculation of latent heat flux. Agricultural and
Forest Meteorology 37, 75–88. doi:10.1016/0168-1923(86)90029-8.
Christopher JT, Manschadi AM, Hammer GL, Borrell AK (2008) Developmental and
physiological traits associated with high yield and stay-green phenotype in wheat. Australian Journal of Agricultural Research 59, 354–364.
Condon AG, Richards RA, Rebetzke GJ, Farquhar GD (2002) Improving intrinsic
water-use efficiency and crop yield. Crop Science 42, 122. doi:10.2135/cropsci2002.0122.
Cowan IR (1965) Transport of water in the soil-plant-atmosphere system. Journal of Applied Ecology 2, 221–239. doi:10.2307/2401706.
Dardanelli JL, Bachmeier OA, Sereno R, Gil R (1997) Rooting depth and soil water
extraction patterns of different crops in a silty loam Haplustoll. Field Crops Research 54, 29–38. doi:10.1016/S0378-4290(97)00017-8.
Dardanelli JL, Ritchie JT, Calmon M, Andriani JM, Collino DJ (2004) An empirical model for root water uptake. Field Crops Research 87, 59–71. doi:10.1016/j.fcr.2003.09.008.
De Wit C, Penning de Vries F (1983) Crop growth models without hormones. Netherlands journal of agricultural science 31, 313–323.
References 211
Deery DM (2008) Water uptake by a single plant: Analysis using experimentation and modeling. Charles Sturt University.
Deery DM, Passioura JB, Condon JR, Katupitiya A (2012) Uptake of water from a
Kandosol subsoil: I. Determination of soil water diffusivity. Plant and Soil 368, 483–492. doi:10.1007/s11104-012-1525-8.
Delhaize E, Ryan PR, Hebb DM, Yamamoto Y, Sasaki T, Matsumoto H (2004) Engineering high-level aluminum tolerance in barley with the ALMT1 gene. Proceedings of the National Academy of Sciences of the United States of
America 101, 15249–15254. doi:10.1073/pnas.0406258101.
Denison RF (2012) ‗Darwinian agriculture: how understanding evolution can improve
agriculture.‘ (Princeton University Press)
Denison RF (2015) Evolutionary tradeoffs as opportunities to improve yield potential. Field Crops Research 182, 3–8. doi:10.1016/j.fcr.2015.04.004.
Depuydt S (2014) Arguments for and against self and non-self root recognition in plants. Functional Plant Ecology 5, 614. doi:10.3389/fpls.2014.00614.
Donald CM (1968) The breeding of crop ideotypes. Euphytica 17, 385–403. doi:10.1007/BF00056241.
Drew MC, Saker LR (1980) Assessment of a rapid method, using soil cores, for
estimating the amount and distribution of crop roots in the field. Plant and Soil 55, 297–305. doi:10.1007/BF02181809.
Duggan BL, Richards RA, van Herwaarden AF (2005) Agronomic evaluation of a tiller inhibition gene (tin) in wheat. II. Growth and partitioning of assimilate. Australian Journal of Agricultural Research 56, 179–186.
Dunin FX, Barrs HD, Meyer WS, Trevitt ACF (1991) Foliage temperature and latent heat flux of irrigated wheat. Agricultural and Forest Meteorology 55, 133–147.
doi:10.1016/0168-1923(91)90027-N.
Elzhov TV, Mullen KM, Spiess A-N, Bolker B (2013) ‗minpack.lm: R interface to the Levenberg-Marquardt nonlinear least-squares algorithm found in MINPACK,
plus support for bounds.‘ http://CRAN.R-project.org/package=minpack.lm.
Evett S, Heng L, Moutonnet P, Nguyen M (2008) Field estimation of soil water content:
A practical guide to methods, instrumentation, and sensor technology. IAEA: Vienna.
Falik O, Reides P, Gersani M, Novoplansky A (2003) Self/non-self discrimination in
roots. Journal of Ecology 91, 525–531. doi:10.1046/j.1365-2745.2003.00795.x.
212 References
Fischer RA (1985) Number of kernels in wheat crops and the influence of solar radiation and temperature. The Journal of Agricultural Science 105, 447–461. doi:10.1017/S0021859600056495.
French R, Schultz J (1984) Water use efficiency of wheat in a Mediterranean-type environment. I. The relation between yield, water use and climate. Australian
Journal of Agricultural Research 35, 743–764.
van Genuchten MT (1980) A closed-form equation for predicting the hydraulic conductivity of unsaturated soils. Soil science society of America journal 44,
892–898.
van Genuchten MT, Leij F, Yates S, others (1991) ‗The RETC code for quantifying the
hydraulic functions of unsaturated soils.‘ (Robert S. Kerr Environmental Research Laboratory)
Gersani M, Brown JS, O‘Brien EE, Maina GM, Abramsky Z (2001) Tragedy of the
commons as a result of root competition. Journal of Ecology 89, 660–669. doi:10.1046/j.0022-0477.2001.00609.x.
Ghezzehei T, Bogie N, Albalasmeh A (2015) Plant roots can actively regulate hydraulic redistribution by modifying the hydraulic properties of the rhizosphere using exudates. In ‗Geophys. Res. Abstr.‘,
Giunta F, Motzo R, Deidda M (1993) Effect of drought on yield and yield components of durum wheat and triticale in a Mediterranean environment. Field Crops
Research 33, 399–409. doi:10.1016/0378-4290(93)90161-F.
Gomez-Macpherson H, Richards RA (1995) Effect of sowing time on yield and agronomic characteristics of wheat in south-eastern Australia. Australian
Journal of Agricultural Research 46, 1381–1399.
Gornall J, Betts R, Burke E, Clark R, Camp J, Willett K, Wiltshire A (2010)
Implications of climate change for agricultural productivity in the early twenty-first century. Philosophical Transactions of the Royal Society of London B: Biological Sciences 365, 2973–2989. doi:10.1098/rstb.2010.0158.
Greenhouse SW, Geisser S (1959) On methods in the analysis of profile data. Psychometrika 24, 95–112. doi:10.1007/BF02289823.
Gruntman M, Novoplansky A (2004) Physiologically mediated self/non-self discrimination in roots. Proceedings of the National Academy of Sciences of the United States of America 101, 3863–3867. doi:10.1073/pnas.0306604101.
Hammer GL, Dong Z, McLean G, Doherty A, Messina C, Schussler J, Zinselmeier C, Paszkiewicz S, Cooper M (2009) Can changes in canopy and/or root system
architecture explain historical maize yield trends in the US corn belt? Crop Science 49, 299. doi:10.2135/cropsci2008.03.0152.
References 213
Hammer GL, Sinclair TR, Chapman SC, Oosterom E van (2004) On systems thinking, systems biology, and the in silico plant. Plant Physiology 134, 909–911. doi:10.1104/pp.103.034827.
Hendriks PW, Kirkegaard JA, Lilley JM, Gregory PJ, Rebetzke GJ (2015) A tillering inhibition gene influences root–shoot carbon partitioning and pattern of water
use to improve wheat productivity in rainfed environments. Journal of Experimental Botany erv457. doi:10.1093/jxb/erv457.
Idso SB (1982) Non-water-stressed baselines: A key to measuring and interpreting plant
water stress. Agricultural Meteorology 27, 59–70. doi:10.1016/0002-1571(82)90020-6.
Idso SB, Jackson RD, Pinter Jr. PJ, Reginato RJ, Hatfield JL (1981) Normalizing the stress-degree-day parameter for environmental variability. Agricultural Meteorology 24, 45–55. doi:10.1016/0002-1571(81)90032-7.
Isbell RF (1993) ‗A Classification System for Australian Soils.‘ (CSIRO Division of Soils)
Johnston WH (2000) Calibration of gypsum blocks and data loggers and their evaluation for monitoring soil water status. Australian Journal of Experimental Agriculture 40, 1131–1136.
Jordan W, Miller F (1980) Genetic variability in sorghum root systems: implications for drought tolerance. Adaptation of Plants to Water and High Temperature Stress
(NC Turner and PJ Kramer, Editors) 383–399.
Keating BA, Meinke H, Probert ME, Huth NI, Hills IG (2001) ‗Nwheat: documentation and performance of a wheat module for APSIM. Tropical Agronomy
Memorandum.‘ (CSIRO Division of Tropical Agriculture)
Kebrom TH, Richards RA (2013) Physiological perspectives of reduced tillering and
stunting in the tiller inhibition (tin) mutant of wheat. Functional Plant Biology 40, 977–985.
Kemanian AR, Stöckle CO, Huggins DR (2005) Transpiration-use efficiency of barley.
Agricultural and Forest Meteorology 130, 1–11. doi:10.1016/j.agrformet.2005.01.003.
Kirkegaard JA, Hunt JR (2010) Increasing productivity by matching farming system management and genotype in water-limited environments. Journal of Experimental Botany erq245. doi:10.1093/jxb/erq245.
Kirkegaard JA, Lilley JM (2007) Root penetration rate -- a benchmark to identify soil and plant limitations to rooting depth in wheat. Australian Journal of
Experimental Agriculture 47, 590–602.
214 References
Kirkegaard JA, Lilley JM, Howe GN, Graham JM (2007) Impact of subsoil water use on wheat yield. Australian Journal of Agricultural Research 58, 303–315.
Klepper B, Belford RK, Rickman RW (1984) Root and Shoot Development in Winter
Wheat1. Agronomy Journal 76, 117. doi:10.2134/agronj1984.00021962007600010029x.
Köppen WP (1918) Klassifikation der klimate nach temperatur, niederschlag und jahreslauf. Petterm Geogr Mitt 64, 193–203, 243–248.
Liao M, Palta JA, Fillery IRP (2006) Root characteristics of vigorous wheat improve
early nitrogen uptake. Australian Journal of Agricultural Research 57, 1097–1107.
Lilley JM, Kirkegaard JA (2011) Benefits of increased soil exploration by wheat roots. Field Crops Research 122, 118–130. doi:10.1016/j.fcr.2011.03.010.
Lopes MS, Reynolds MP (2010) Partitioning of assimilates to deeper roots is associated
with cooler canopies and increased yield under drought in wheat. Functional Plant Biology 37, 147–156.
Ludlow M, Muchow R (1990) A critical evaluation of traits for improving crop yields in water-limited environments. Advances in agronomy 43, 107–153.
Lynch JP (1995) Root architecture and plant productivity. Plant Physiology 109, 7–13.
Lynch JP (2013) Steep, cheap and deep: an ideotype to optimize water and N acquisition by maize root systems. Annals of Botany 112, 347–357.
doi:10.1093/aob/mcs293.
Manschadi AM, Christopher J, deVoil P, Hammer GL (2006) The role of root architectural traits in adaptation of wheat to water-limited environments.
Functional Plant Biology 33, 823–837.
Manschadi AM, Hammer GL, Christopher JT, deVoil P (2008) Genotypic variation in
seedling root architectural traits and implications for drought adaptation in wheat (Triticum aestivum L.). Plant and Soil 303, 115–129. doi:10.1007/s11104-007-9492-1.
Miralles DJ, Slafer GA, Lynch V (1997) Rooting patterns in near-isogenic lines of spring wheat for dwarfism. Plant and Soil 197, 79–86.
doi:10.1023/A:1004207407979.
Mitchell JH, Chapman SC, Rebetzke GJ, Bonnett DG, Fukai S (2012) Evaluation of a reduced-tillering (tin) gene in wheat lines grown across different production
environments. Crop and Pasture Science 63, 128–141.
References 215
Murphy SR, Lodge GM (2004) Surface soil water dynamics in pastures in northern New South Wales. 1. Use of electrical resistance sensors. Australian Journal of Experimental Agriculture 44, 273–281.
Northcote KH (1971) A factual key for the recognition of Australian soils.
Olivares-Villegas JJ, Reynolds MP, McDonald GK (2007) Drought-adaptive attributes
in the Seri/Babax hexaploid wheat population. Functional Plant Biology 34, 189–203.
Palta JA, Chen X, Milroy SP, Rebetzke GJ, Dreccer MF, Watt M (2011) Large root
systems: are they useful in adapting wheat to dry environments? Functional Plant Biology 38, 347–354.
Palta JA, Watt M (2009) Chapter 13 - Vigorous crop root systems: form and function for improving the capture of water and nutrients. ‗Crop Physiol.‘ (Eds VO Sadras, D Calderini) pp. 309–325. (Academic Press: San Diego)
http://www.sciencedirect.com/science/article/pii/B978012374431900013X.
Pask AJD, Reynolds MP (2013) Breeding for Yield Potential has Increased Deep Soil
Water Extraction Capacity in Irrigated Wheat. Crop Science 53, 2090. doi:10.2135/cropsci2013.01.0011.
Passioura JB (1976) Physiology of grain yield in wheat growing on stored water.
Functional Plant Biology 3, 559–565.
Passioura JB (1977) Grain yield, harvest index, and water use of wheat. Journal of the
Australian Institute of Agricultural Science.
Passioura JB (1983) Roots and drought resistance. Agricultural Water Management 7, 265–280. doi:10.1016/0378-3774(83)90089-6.
Passioura JB (2007) The drought environment: physical, biological and agricultural perspectives. Journal of Experimental Botany 58, 113–117.
doi:10.1093/jxb/erl212.
Paydar Z, Cresswell H (1996) Water retention in Australian soils. 2.* Prediction using particle size, bulk density, and other properties. Soil Research 34, 679–693.
Pinto RS, Reynolds MP (2015) Common genetic basis for canopy temperature depression under heat and drought stress associated with optimized root
distribution in bread wheat. Theoretical and Applied Genetics 128, 575–585. doi:10.1007/s00122-015-2453-9.
R Core Team (2014) ‗R: A Language and Environment for Statistical Computing.‘ (R
Foundation for Statistical Computing: Vienna, Austria) http://www.R-project.org/.
216 References
Ray DK, Mueller ND, West PC, Foley JA (2013) Yield trends are insufficient to double global crop production by 2050. PLoS ONE 8, e66428. doi:10.1371/journal.pone.0066428.
Rebetzke GJ, Botwright TL, Moore CS, Richards RA, Condon AG (2004) Genotypic variation in specific leaf area for genetic improvement of early vigour in wheat.
Field Crops Research 88, 179–189. doi:10.1016/j.fcr.2004.01.007.
Rebetzke GJ, Rattey AR, Farquhar GD, Richards RA, Condon A (Tony) G (2012) Genomic regions for canopy temperature and their genetic association with
stomatal conductance and grain yield in wheat. Functional Plant Biology 40, 14–33.
Rebetzke GJ, Richards RA (1999) Genetic improvement of early vigour in wheat. Australian Journal of Agricultural Research 50, 291–302.
Reynolds M, Dreccer F, Trethowan R (2007) Drought-adaptive traits derived from
wheat wild relatives and landraces. Journal of Experimental Botany 58, 177–186. doi:10.1093/jxb/erl250.
Richards RA (1991) Crop improvement for temperate Australia: Future opportunities. Field Crops Research 26, 141–169. doi:10.1016/0378-4290(91)90033-R.
Richards RA (1992) The effect of dwarfing genes in spring wheat in dry environments.
II. Growth, water use and water-use efficiency. Australian Journal of Agricultural Research 43, 529–539.
Richards RA, Passioura JB (1989) A breeding program to reduce the diameter of the major xylem vessel in the seminal roots of wheat and its effect on grain yield in rain-fed environments. Australian Journal of Agricultural Research 40, 943–
950.
Richards RA, Watt M, Rebetzke GJ (2006) Physiological traits and cereal germplasm
for sustainable agricultural systems. Euphytica 154, 409–425. doi:10.1007/s10681-006-9286-1.
Sadras VO (2007) Evolutionary aspects of the trade-off between seed size and number
in crops. Field Crops Research 100, 125–138. doi:10.1016/j.fcr.2006.07.004.
Sadras VO, Connor DJ (1991) Physiological basis of the response of harvest index to
the fraction of water transpired after anthesis: A simple model to estimate harvest index for determinate species. Field Crops Research 26, 227–239. doi:10.1016/0378-4290(91)90001-C.
Sadras VO, Lawson C (2013) Nitrogen and water-use efficiency of Australian wheat varieties released between 1958 and 2007. European Journal of Agronomy 46,
34–41. doi:10.1016/j.eja.2012.11.008.
References 217
Sadras VO, Milroy SP (1996) Soil-water thresholds for the responses of leaf expansion and gas exchange: A review. Field Crops Research 47, 253–266. doi:10.1016/0378-4290(96)00014-7.
Semchenko M, Saar S, Lepik A (2014) Plant root exudates mediate neighbour recognition and trigger complex behavioural changes. New Phytologist 204,
631–637. doi:10.1111/nph.12930.
Siddique K, Tennant D, Perry M, Belford R (1990) Water use and water use efficiency of old and modern wheat cultivars in a Mediterranean-type environment.
Australian Journal of Agricultural Research 41, 431–447.
Sinclair TR, Hammer GL, van Oosterom EJ (2005) Potential yield and water-use
efficiency benefits in sorghum from limited maximum transpiration rate. Functional Plant Biology 32, 945–952.
Sirault XR (2007) Leaf rolling in wheat. Australian National University.
Sleeman JR (1979) The Soils of the Ginninderra Experiment Station, ACT. CSIRO, 41.
Smucker A, McBurney S, Srivastava A (1982) Quantitative separation of roots from
compacted soil profiles by the hydropneumatic elutriation system. Agronomy Journal 74, 500–503.
Soil Survey Staff (1999) ‗Soil taxonomy: a basic system of soil classification for
making and interpreting soil surveys.‘ (US Department of Agriculture, Soil Conservation Service)
Stace HC, Hubble GD, Brewer R, Northcote KH, Sleeman JR, Mulcahy MJ, Hallsworth EG (1968) A handbook of Australian soils. https://trid.trb.org/view.aspx?id=1199558.
Stirzaker RJ, Passioura JB (1996) The water relations of the root–soil interface. Plant, Cell & Environment 19, 201–208. doi:10.1111/j.1365-3040.1996.tb00241.x.
Tennant D (1976) Root growth of wheat. I. Early patterns of multiplication and extension of wheat roots including effects of levels of nitrogen, phosphorus and potassium. Crop and Pasture Science 27, 183–196.
Trachsel S, Kaeppler SM, Brown KM, Lynch JP (2013) Maize root growth angles become steeper under low N conditions. Field Crops Research 140, 18–31.
doi:10.1016/j.fcr.2012.09.010.
Wasson AP, Rebetzke GJ, Kirkegaard JA, Christopher J, Richards RA, Watt M (2014) Soil coring at multiple field environments can directly quantify variation in deep
root traits to select wheat genotypes for breeding. Journal of Experimental Botany. doi:10.1093/jxb/eru250.
218 References
Wasson AP, Richards RA, Chatrath R, Misra SC, Prasad SVS, Rebetzke GJ, Kirkegaard JA, Christopher J, Watt M (2012) Traits and selection strategies to improve root systems and water uptake in water-limited wheat crops. Journal of Experimental
Botany 63, 3485–3498. doi:10.1093/jxb/ers111.
Watt M, Kirkegaard JA, Rebetzke GJ (2005) A wheat genotype developed for rapid leaf
growth copes well with the physical and biological constraints of unploughed soil. Functional Plant Biology 32, 695–706.
Watt M, Magee LJ, McCully ME (2008) Types, structure and potential for axial water
flow in the deepest roots of field-grown cereals. New Phytologist 178, 135–146. doi:10.1111/j.1469-8137.2007.02358.x.
Watt M, Moosavi S, Cunningham SC, Kirkegaard JA, Rebetzke GJ, Richards RA (2013) A rapid, controlled-environment seedling root screen for wheat correlates well with rooting depths at vegetative, but not reproductive, stages at two field
sites. Annals of Botany 112, 447–455. doi:10.1093/aob/mct122.
Wesseling J, van Wijk WR, Fireman M, van‘t Woudt BD, Hagan RM (1957) ‗Land
drainage in relation to soils and crops in drainage of agricultural lands.‘ (American Society of Agronomy) https://dl.sciencesocieties.org/publications/books/abstracts/agronomymonogra/dr
ainageofagric/461.
White RG, Kirkegaard JA (2010) The distribution and abundance of wheat roots in a
dense, structured subsoil – implications for water uptake. Plant, Cell & Environment 33, 133–148. doi:10.1111/j.1365-3040.2009.02059.x.
Wikipedia (2014) Köppen climate classification --- Wikipedia, The Free Encyclopedia.
http://en.wikipedia.org/w/index.php?title=K%C3%B6ppen_climate_classification&oldid=606957479.
Wilson PB, Rebetzke GJ, Condon AG (2015) Pyramiding greater early vigour and integrated transpiration efficiency in bread wheat; trade-offs and benefits. Field Crops Research 183, 102–110. doi:10.1016/j.fcr.2015.07.002.
Wojciechowski T, Gooding M, Ramsay L, Gregory P (2009) The effects of dwarfing genes on seedling root growth of wheat. Journal of Experimental Botany erp107.
Zadoks JC, Chang TT, Konzak CF (1974) A decimal code for the growth stages of cereals. Weed Research 14, 415–421. doi:10.1111/j.1365-3180.1974.tb01084.x.
Zhang D-Y, Sun G-J, Jiang X-H (1999) Donald‘s ideotype and growth redundancy: a
game theoretical analysis. Field Crops Research 61, 179–187. doi:10.1016/S0378-4290(98)00156-7.
Zotarelli L, Dukes MD, Romero CC, Migliaccio KW, Morgan KT (2010) Step by step calculation of the Penman-Monteith Evapotranspiration (FAO-56 Method).
References 219